Suse
Monthly
Local privilege escalation in the kernel-mode layer of the NVIDIA GPU Display Driver for Linux allows an unprivileged local user to bypass read-only memory protection and gain write access to memory the driver marked read-only, which can result in kernel code execution, denial of service, privilege escalation, information disclosure, and data tampering. Exploitation requires the attacker to already hold low-privileged local access on an affected Linux host (CVSS 3.1 vector AV:L/AC:L/PR:L/UI:N, score 7.8) with no user interaction and no network vector; the risk is greatest on shared, multi-tenant vGPU, VDI, and cloud-GPU hosts where crossing the user-to-kernel boundary breaks tenant isolation. No public exploit code has been identified at time of analysis and the issue is not listed in CISA KEV, but NVIDIA has published a fix through security bulletin 5861. This is a genuine local privilege-escalation risk rather than an emergency, and it should be prioritized on multi-tenant GPU infrastructure.
Local privilege escalation in NVIDIA Display Driver for Windows and Linux allows an unprivileged local attacker to trigger a use-after-free condition by issuing a crafted sequence of driver commands, potentially achieving code execution, privilege escalation, denial of service, or information disclosure. The vulnerability affects systems with GeForce, RTX/Quadro/NVS, or Tesla GPUs running vulnerable driver versions. No public exploit identified at time of analysis.
Local privilege escalation and memory corruption in NVIDIA GPU Display Driver on Windows and Linux, affecting GeForce, RTX/Quadro/NVS, and Tesla lines. An unprivileged local user can trigger a use-after-free (CWE-416) by issuing a specific sequence of driver commands, leading to potential code execution, privilege escalation, denial of service, or information disclosure. Requires local access and low privileges (AV:L/PR:L) with no user interaction (UI:N); no public exploit identified at time of analysis. Risk is highest in multi-user or shared-GPU environments such as VDI, terminal servers, and cloud/virtualized GPU hosts.
Local privilege escalation in the NVIDIA GPU Display Driver for Linux allows an authenticated, unprivileged user to trigger a use-after-free condition by issuing a specific sequence of driver commands, potentially leading to code execution, privilege escalation, denial of service, and information disclosure. The vulnerability affects GeForce, RTX/Quadro/NVS, and Tesla GPU drivers, including guest-VM vGPU deployments, and requires local access with low privileges. No public exploit code or confirmed active exploitation has been identified at time of analysis.
Use-after-free in the NVIDIA GPU Display Driver for Linux allows a local, unprivileged user to corrupt driver memory by racing or abusing object lifetime handling, with a successful exploit potentially yielding code execution in kernel context, privilege escalation to root, denial of service, information disclosure, and data tampering. The flaw scores 7.8 (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) and is confined to local attackers who already hold an account on the GPU host, with no network vector and no user interaction required. This is a genuine but not emergency-grade issue: no public exploit code has been identified at time of analysis and the CVE is not listed in CISA KEV, so realistic exposure is concentrated on shared or multi-tenant GPU systems - VDI desktops, cloud GPU instances, and HPC nodes - where untrusted users share a single driver instance.
Local code execution and denial of service on Windows systems with affected NVIDIA GPU Display Drivers stem from an integer overflow in the display driver's DIAG escape handler that results in an out-of-bounds write (CWE-787). An unprivileged local user (CVSS PR:L) can trigger the flaw via a device I/O call to the handler; no user interaction is required, and exploitation is confined to the local machine, meaning an attacker must already have a foothold. No public exploit identified at time of analysis (and no CISA KEV entry), but the shared-system impact-multi-user workstations, VDI, and cloud GPU instances-raises the risk profile.
Type confusion in the kernel-mode layer of the NVIDIA GPU Display Driver for Linux allows a local low-privileged user (CVSS vector AV:L/AC:L/PR:L/UI:N) who can reach the driver's device-node/ioctl interface to trigger memory corruption with high impact across confidentiality, integrity, and availability, potentially yielding code execution in kernel context, privilege escalation to root, denial of service, information disclosure, and data tampering. The CVSS 3.1 base score of 7.8 reflects a local attack vector paired with uniformly high impacts rather than any remote reachability, and despite the 'RCE' tag in the source metadata the vulnerability is not remotely exploitable. No public exploit code or confirmed in-the-wild exploitation was identified at time of analysis, and no EPSS or CISA KEV signal was provided.
A type-confusion flaw in the kernel-mode layer of the NVIDIA GPU Display Driver for Linux lets a local, low-privileged user who can open the driver's device nodes and issue ioctls trigger memory corruption by winning a handle-recycle race, yielding kernel-level code execution, privilege escalation, denial of service, information disclosure and data tampering. Affected platforms are the Linux driver branches covering GeForce, RTX/Quadro/NVS, Tesla/data-center GPUs and the vGPU guest driver and Virtual GPU Manager (CVSS 3.1 base 7.8; our independent assessment rates it AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H, i.e. 7.0, because the race window makes exploitation timing-sensitive rather than deterministic). This is a local privilege-escalation-class issue with no remote vector: no public exploit code was identified at time of analysis and the flaw is not confirmed actively exploited (CISA KEV), though it is most consequential on multi-user, VDI and cloud/virtualized GPU hosts where untrusted tenants share a single driver instance.
Local privilege escalation in NVIDIA GPU Display Driver for Linux allows an unprivileged local user to trigger an out-of-bounds write in kernel memory via a crafted ioctl call with an unvalidated buffer size, enabling code execution, privilege escalation, information disclosure, and denial of service. The flaw affects GeForce, RTX/Quadro/NVS, and Tesla drivers, as well as Guest Driver and Virtual GPU Manager components, with highest risk in multi-tenant GPU environments such as VDI and cloud/HPC deployments. No public exploit identified at time of analysis; CVSS 3.1 score is 7.8 (local, low privileges, no user interaction).
Local privilege escalation in the NVIDIA GPU Display Driver for Linux allows an authenticated unprivileged user to trigger a use-after-free in the kernel mode layer, potentially achieving code execution, denial of service, privilege escalation, information disclosure, and data tampering. The flaw is reachable only by a locally authenticated low-privileged account (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) that can issue ioctls to the GPU kernel-mode device interface; there is no remote attack path and no user interaction is required once a foothold exists. The 'Guest Driver' classification makes the issue most consequential in virtualized/vGPU and multi-tenant datacenter environments, where a low-privileged guest can reach the shared GPU driver. No public exploit identified at time of analysis, and the vendor advisory is the authoritative source for fixed driver versions.
Local privilege escalation in NVIDIA's GPU Display Driver (GeForce, RTX/Quadro/NVS, and Tesla, on both Windows and Linux) allows an authenticated low-privileged local user to reach GPU memory belonging to another process, enabling code execution, escalation of privileges, data tampering, denial of service, and information disclosure. Per the assessed vector (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) there is no remote or network path; the flaw is most consequential on shared, multi-user, VDI, or multi-tenant GPU hosts where cross-process GPU memory isolation is the boundary between users and tenants. No public exploit code and no confirmed active exploitation were identified at time of analysis, and the provided vendor reference points to NVIDIA's product security advisory (2026/5861) rather than a fixed driver version named in this data set.
Local privilege escalation in the NVIDIA GPU Display Driver for Linux kernel mode layer allows an authenticated, unprivileged user to trigger a double-free of imported memory state, potentially leading to code execution, denial of service, information disclosure, and data tampering. The flaw requires local access and a low-privilege account; no public exploit code or active exploitation has been identified at time of analysis. The CVSS 3.1 score is 7.8 (HIGH).
Local privilege escalation, code execution, denial of service, and information disclosure in NVIDIA GPU Display Driver on Windows and Linux stem from an integer overflow (CWE-190) in the kernel-mode layer that an unprivileged user can trigger to cause an out-of-bounds write. Exploitation requires local access with an existing low-privilege account (authenticated, PR:L) on a host with an affected GeForce, RTX/Quadro/NVS, or Tesla driver; the CVSS v3.1 base score is 7.8 (High) with a local attack vector (AV:L/AC:L/PR:L/UI:N). No public exploit code or CISA KEV entry was identified at the time of analysis, but the flaw is genuine and actionable for local privilege escalation if the vulnerable driver interface is reachable.
An out-of-bounds write in the kernel-mode layer of the NVIDIA GPU Display Driver for Windows and Linux allows an unprivileged local user to corrupt kernel memory, which can lead to code execution in the kernel, privilege escalation, denial of service, information disclosure, or data tampering. Systems running GeForce, RTX/Quadro/NVS, or Tesla driver packages on either operating system are affected; per the CVSS vector (AV:L, PR:L, UI:N) the flaw is reachable only by a user who already has local code execution on the host, making it a local privilege-escalation primitive rather than a remote entry point. No public exploit has been identified at time of analysis.
Local privilege escalation in NVIDIA GPU Display Driver for Linux stems from a missing authorization check in the kernel mode layer, allowing an unprivileged local user to modify privileged configuration and potentially achieve code execution, escalation of privileges, denial of service, information disclosure, and data tampering. Exploitation requires a pre-existing local low-privileged foothold (CVSS:3.1/AV:L/AC:L/PR:L/UI:N), such as an unprivileged account on a host or inside a guest VM using the NVIDIA guest driver or Virtual GPU Manager, so this is an escalation primitive rather than a remote initial-access vector. No public exploit code is identified and no CISA KEV active exploitation is confirmed at time of analysis, but the high confidentiality, integrity, and availability impacts and broad affected driver families make patching a priority.
Use-after-free in the NVIDIA GPU Display Driver kernel mode layer (Windows and Linux) lets a local user who already holds low-privileged code execution on the host trigger a memory-corruption condition inside the kernel driver, with the potential for code execution, privilege escalation, information disclosure, data tampering, and denial of service. Affected platforms span consumer and professional GPU lines (GeForce, RTX/Quadro/NVS, Tesla) as well as the vGPU components, where a malicious or compromised guest can exercise the vulnerable path through the Guest Driver and Virtual GPU Manager. The attack is strictly local (CVSS:3.1/AV:L/AC:L/PR:L/UI:N, score 7.8), requires no user interaction and no non-default configuration beyond having the driver installed, and there is no public exploit identified at time of analysis - no KEV entry and no proof-of-concept code is referenced in the available data.
A locally-triggered out-of-bounds write in the NVIDIA GPU Display Driver kernel-mode layer lets an attacker who already holds low-level privileges on a Windows or Linux host corrupt kernel memory, with the potential for code execution in kernel context, escalation of privileges, denial of service, information disclosure and data tampering. The affected driver ships by default across the GeForce, RTX/Quadro/NVS, Tesla and Virtual GPU Manager product lines, so any host with a current NVIDIA driver installed is in scope, but exploitation is not remote: an attacker must first obtain code execution on the target (an authenticated local user, a compromised low-privileged process, or sandboxed code with driver access) and then develop a reliable kernel corruption exploit. CVE-2026-47548 carries a vendor/NVD CVSS 3.1 vector of AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8, High); no public exploit code has been identified at time of analysis and the CVE is not listed in CISA KEV.
Local attackers with low-privileged code execution can trigger an out-of-bounds read (CWE-125) in the kernel-mode layer of the NVIDIA GPU Display Driver on Windows and Linux, which may lead to escalation of privileges, code execution, denial of service, information disclosure, and data tampering. This locally exploitable flaw (CVSS 7.8, AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) affects GeForce, RTX/Quadro/NVS, Tesla, and Virtual GPU Manager products, and no public exploit has been identified at the time of analysis. A vendor advisory with remediation guidance is available from NVIDIA via its product security page.
Integer underflow in the kernel-mode layer of NVIDIA GPU Display Driver for Windows and Linux allows a local attacker with low privileges to potentially achieve code execution, denial of service, privilege escalation, information disclosure, and data tampering. The flaw affects GeForce, RTX/Quadro/NVS, Tesla, and vGPU Manager products, and exploitation requires local access (CVSS AV:L) with no user interaction (UI:N) but low privileges (PR:L). No public exploit code has been identified at time of analysis, and there is no indication of active exploitation.
An out-of-bounds write in the kernel-mode layer of the NVIDIA GPU Display Driver for Windows and Linux allows a local, low-privileged authenticated user to corrupt kernel memory, with the potential to escalate privileges to kernel-level code execution, crash the system (denial of service), disclose information, or tamper with data. The flaw rates CVSS 3.1 7.8 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H), meaning exploitation requires pre-existing code execution as a standard user on a host with the driver loaded, but no user interaction and no non-default configuration; the vulnerable kernel-mode layer ships and loads by default once the driver is installed. There is no confirmed active exploitation (not in CISA KEV) and no public exploit identified at time of analysis, but the affected GeForce, RTX/Quadro/NVS, and Tesla product families are extremely widely deployed, making this a genuine high-priority patch-management item rather than an emergency.
Local attackers with low privileges can exploit an uninitialized pointer access flaw (CWE-824) in the firmware of NVIDIA Display Driver for Windows and Linux to achieve code execution, denial of service, privilege escalation, information disclosure, or data tampering. The vulnerability requires local access (AV:L) and authenticated low-privilege credentials (PR:L); no public exploit identified at time of analysis. Affected products include GeForce, RTX/Quadro/NVS, and Tesla product lines, including vGPU Manager deployments.
A use-after-free (CWE-416) in the NVIDIA GPU Display Driver for Windows and Linux lets a local unprivileged user free and then re-reference a driver-managed memory object, potentially yielding code execution in kernel context, privilege escalation, data tampering, denial of service, and information disclosure. The flaw is rated CVSS 3.1 7.8 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) and affects GeForce, RTX/Quadro/NVS, and Tesla product lines; the 'Guest Driver' tag indicates the vulnerable path may also be reachable from inside a virtualized guest, making guest-to-host or guest-context escalation a relevant scenario. Exploitation requires that the attacker already hold a local foothold with normal user privileges on a host with the affected driver installed and the GPU device interface exposed - there is no remote/network vector - and at time of analysis no public exploit code and no confirmed active exploitation have been identified, so this is a priority local privilege-escalation issue rather than an imminent, opportunistically exploited threat.
Local kernel stack memory disclosure in NVIDIA GPU Display Driver for Windows and Linux allows an attacker who already has low-privilege authenticated code execution on the host to read kernel stack contents such as return addresses and pointers. Affected installations include GeForce, RTX/Quadro/NVS, and Tesla driver configurations, and the independent assessment rates the flaw as AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N, meaning high confidentiality impact without confirmed integrity or availability impact. No public exploit is identified at time of analysis, and no CISA KEV active exploitation is reported.
An out-of-bounds read in the kernel-mode layer of the NVIDIA GPU Display Driver for Windows and Linux allows a local, low-privileged user to read beyond the bounds of kernel heap memory, producing information disclosure and denial of service, with potential for privilege escalation, code execution, and data tampering if the read primitive is successfully leveraged. The flaw affects GeForce, RTX/Quadro/NVS, and Tesla product families on systems with the vulnerable driver branch installed, and exploitation requires an attacker to already hold local access to the host (per the assessed CVSS:3.1/AV:L/AC:L/PR:L/UI:N vector, base 7.8), with no user interaction needed. No public exploit code has been identified at time of analysis, and the issue is not listed in CISA KEV, so exploitation is currently expected to require a motivated local attacker rather than being broadly weaponized.
An out-of-bounds read in the kernel-mode layer of the NVIDIA GPU Display Driver for Windows and Linux lets a local attacker who already holds a low-privileged execution context on the host read past the bounds of a kernel buffer, exposing kernel memory contents and creating a pathway toward further privilege escalation or data tampering. Exploitation is constrained to the local attack surface: the attacker must already be able to run code on the machine (AV:L, PR:L) and no user interaction or network path is involved, so despite the 7.8 headline score the practical risk is a local privilege/information-disclosure concern rather than a remotely reachable, mass-exploitable flaw. There is no confirmed active exploitation (not listed in CISA KEV) and no public exploit code identified at time of analysis; the affected hardware spans GeForce, RTX/Quadro/NVS, and Tesla products, with an NVIDIA vendor advisory available.
Local privilege escalation in the NVIDIA GPU Display Driver kernel-mode layer allows a user with standard privileges to trigger an out-of-bounds write in kernel memory, potentially achieving kernel-level code execution, denial of service, information disclosure and data tampering. The affected driver branches span NVIDIA GeForce, RTX/Quadro/NVS and Tesla families on both Windows and Linux, and the 'Guest Driver' tag indicates virtualized/vGPU and guest-VM driver deployments are also in scope, which matters for GPU-accelerated VDI and datacenter estates. Exploitation is constrained to the host: the CVSS 3.1 vector is AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8), so an attacker must already have a foothold and the ability to run code as a low-privileged local user; no public exploit code and no confirmed active exploitation were identified at time of analysis.
Local privilege escalation in NVIDIA GPU Display Driver for Windows and Linux allows a low-privileged local user to trigger an integer overflow in the kernel mode layer, resulting in an out-of-bounds write to GPU memory that can lead to code execution, denial of service, escalation of privileges, information disclosure, and data tampering. Affected platforms include GeForce, RTX/Quadro/NVS, and Tesla drivers, with the 'Guest Driver' tag indicating that VM guests with GPU access are also in scope for guest-to-host or in-guest escalation. No public exploit code or confirmed active exploitation (CISA KEV) has been identified at time of analysis, and the CVSS 3.1 score is 7.8 (AV:L/PR:L/UI:N).
Local privilege escalation in NVIDIA GPU Display Driver for Windows and Linux allows a low-privileged user to trigger an incorrect conversion between numeric types in the kernel mode layer, potentially leading to code execution, denial of service, escalation of privileges, information disclosure, and data tampering. The vulnerability requires local access and low-level privileges (CVSS vector PR:L) with no user interaction, and it is not remotely exploitable. No public exploit code has been identified at time of analysis, but the risk is materially higher on shared or multi-tenant GPU hosts such as VDI, vGPU, or Tesla datacenter deployments where untrusted users can reach the driver.
Local privilege escalation in the NVIDIA GPU Display Driver for Windows and Linux lets a low-privileged local user trigger an out-of-bounds array access in the kernel mode layer, potentially yielding code execution in kernel context, denial of service, escalation of privileges, information disclosure and data tampering. The assessed CVSS 3.1 vector is AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8), meaning the attack is strictly local and requires only low-privileged authenticated code on the host, with no user interaction and low complexity - so any unprivileged process able to reach the affected kernel-mode interface on a system running GeForce, RTX/Quadro/NVS or Tesla drivers (or a vGPU host reachable from a guest) can attempt it. There is confirmed actively exploited (CISA KEV) status: no - the CVE is not in KEV and no public exploit code has been identified at time of analysis, so early patching is key while exploitation activity has not yet been observed.
Local code execution, privilege escalation, denial of service, information disclosure, and data tampering are possible on Linux systems running vulnerable NVIDIA GPU Display Drivers (GeForce, RTX/Quadro/NVS, and Tesla branches) because of a type-confusion flaw (CWE-843) in the outdated embedded cryptographic library used by the NGX updater. The assessed profile is CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8): an attacker who already holds a low-privileged local account can reach the vulnerable updater code path, with no user interaction required and low complexity, but the flaw is not remotely reachable, so shared and multi-user hosts (VDI, HPC/Tesla compute clusters, and workstations with untrusted local users) are the realistic risk surface. No public exploit code has been identified at time of analysis, and the exact affected and fixed driver versions must be confirmed from NVIDIA's advisory.
Integer overflow in the kernel-mode layer of NVIDIA vGPU Virtual GPU Manager can be triggered by a low-privileged user inside a guest VM, causing memory corruption with high confidentiality, integrity, and availability impact. The vulnerability affects vGPU deployments on Windows and Linux hosts across GeForce, RTX/Quadro/NVS, and Tesla driver branches, and successful exploitation could enable code execution, denial of service, privilege escalation, information disclosure, or data tampering. No public exploit code or active exploitation has been identified in the provided intelligence at time of analysis.
Local privilege escalation in the NVIDIA GPU Display Driver for Linux allows an authenticated, low-privileged user to drive an out-of-bounds write in the driver's kernel-mode layer by supplying mismatched memory buffers during event buffer setup, with a successful exploit potentially leading to kernel code execution, denial of service, privilege escalation, information disclosure, and data tampering (CVSS 3.1 base 7.8, AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H). The affected code is the Linux kernel-mode component shipped with GeForce, RTX/Quadro/NVS and Tesla driver packages, and the same path can be reached by untrusted code inside a guest VM using the NVIDIA vGPU guest driver on a host running the Virtual GPU Manager. No public exploit code was identified and there is no confirmed active exploitation (CISA KEV) at time of analysis; the bug is local-only, requiring the attacker to already run code with at least low privileges on the host or in a guest, and it is not remotely reachable.
Local privilege escalation in the NVIDIA GPU Display Driver kernel-mode layer lets an attacker who already holds low-privileged code execution on a Windows or Linux host - including code running inside a vGPU guest VM - drive the driver into an error path that mishandles reference counts, producing a use-after-free. The corrupted object is kernel memory, so successful exploitation can yield code execution, escalation of privileges, denial of service, information disclosure, and data tampering (CVSS 7.8, AV:L/AC:L/PR:L/UI:N); it is not remotely reachable and requires a vulnerable driver version to be installed. The GeForce, RTX/Quadro/NVS, and Tesla driver families are in scope, and no public exploit code has been identified at the time of analysis.
Local privilege escalation in the NVIDIA GPU Display Driver for Linux lets an already-authenticated, low-privileged user trigger a format string defect (CWE-134) and potentially achieve code execution, privilege escalation, data tampering, denial of service, and information disclosure; NVIDIA rates it 7.8 (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H). The attack surface is strictly local - there is no remote vector - so real risk concentrates on multi-user and multi-tenant GPU systems such as HPC clusters, ML/AI training nodes, shared engineering workstations, and cloud GPU instances, where a low-privileged tenant could reach the driver while other users' data and the host remain in scope. No public exploit code has been identified at time of analysis and the CVE is not listed in CISA KEV; the impact rating of total confidentiality, integrity, and availability loss is consistent with memory corruption inside a privileged driver context.
Local privilege escalation and kernel memory-safety compromise on Linux systems running NVIDIA GPU Display Drivers allow an unprivileged local user to cause improper release of kernel-mode memory resources, leaving a mapping accessible after the underlying memory has been reused. The flaw affects GeForce, RTX/Quadro/NVS, and Tesla driver lines and carries a CVSS 3.1 base score of 7.8 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H), meaning an attacker who already holds a local login can potentially achieve code execution, denial of service, information disclosure, privilege escalation, and data tampering without user interaction. No public exploit has been identified at time of analysis and the issue is not confirmed as actively exploited, but the local-only vector makes it most dangerous on shared or multi-tenant GPU hosts where one unprivileged tenant could compromise the kernel and other workloads.
Local privilege escalation and memory-protection bypass in NVIDIA GPU Display Driver for Linux allows an unprivileged local user to execute code at higher privilege, tamper with data, disclose information, or cause denial of service. The flaw is in the kernel-mode layer and affects GeForce, RTX/Quadro/NVS, Tesla, and vGPU Guest Driver/Virtual GPU Manager deployments, with a CVSS 3.1 score of 7.8 (AV:L/AC:L/PR:L/UI:N). No public exploit identified at time of analysis, and no CISA KEV active-exploitation signal is present; the assessed attack path requires prior local access on the host or inside a vGPU guest.
Container escape in CRI-O - the OCI runtime that underpins Red Hat OpenShift Container Platform 4 - is achievable when a workload-level actor can create or influence pod metadata that overwrites CRI-O's own reserved sandbox bookkeeping; the poisoned state is then reloaded as trusted after a CRI-O daemon restart, and a subsequent container recreate inside that sandbox can expose a host-side runtime-management resource to the container. Rated CVSS 3.1 7.8 (AV:L/AC:H/PR:L/UI:N/S:C/C:H/I:H/A:H), the flaw is not remotely reachable without cluster workload access, and the mandatory restart-then-recreate sequencing is the principal limiter on realistic exploitation. No public exploit code has been identified at time of analysis and the issue is not confirmed actively exploited (no CISA KEV listing); no EPSS score was supplied in the available intelligence.
Memory corruption in U-Boot's DHCPv6 netboot path allows an attacker on the same local network to crash or corrupt the bootloader by sending crafted DHCPv6 ADVERTISE or REPLY packets with oversized DUID (SERVERID/CLIENTID) option lengths. All U-Boot builds prior to 2026.10-rc5 that perform DHCPv6-based network boot are affected; the assessed impact is availability-focused (CVSS:3.1 AV:A/AC:L/PR:N/UI:N/C:N/I:L/A:H), with confidentiality unaffected and integrity impact limited. Exploitation requires adjacency to the booting device and only works during the brief netboot window, so devices that boot from local storage are not exposed; no public exploit and no confirmed active exploitation were identified at time of analysis.
A use-after-free in U-Boot's lwIP-based wget HTTP client lets an attacker who controls (or MITMs) the server a device downloads from crash the bootloader during a netboot/provisioning fetch, interrupting boot. The bug sits in the httpc_recv_cb() receive callback: when HTTP data storage fails, it frees the connection's protocol control block but returns ERR_BUF instead of ERR_ABRT, so the TCP input path continues on released memory. All U-Boot versions before 2026.10-rc5 are affected; the CVSS 4.0 vector scores 8.8 (AV:N/AC:L/PR:N/UI:N, VA:H, VI:L) and our independent CVSS 3.1 assessment rates it 7.5 as an availability-dominant issue - no public exploit identified at time of analysis, and impact is limited to a bootloader crash rather than code execution or data exposure.
Memory corruption in U-Boot’s NFS READLINK handling can crash the bootloader on devices booting over NFS. A malicious or man-in-the-middle NFS server can return a crafted symlink length that triggers a signed-to-unsigned conversion error (CWE-195) and out-of-bounds access in nfs_readlink_reply(), affecting U-Boot versions before 2026.10-rc5; the vendor CVSS 4.0 base score is 8.8. Exploitation requires the device to be actively performing an NFS netboot and reaching the attacker-controlled server, so idle or non-NFS-boot devices are unaffected, and no public exploit code or CISA KEV active exploitation is identified at time of analysis.
Memory corruption in the U-Boot bootloader's NFS client lets a malicious or man-in-the-middle NFS server overflow the packet buffer in nfs_read_reply(), crashing the bootloader or corrupting memory before control is handed to the kernel. Affected are U-Boot builds prior to 2026.10-rc5 that perform NFS netboot (CPE cpe:2.3:a:u-boot:u-boot:*); the flaw is reachable by an unauthenticated network attacker (CVSS vector PR:N), though per our assessment the practical outcome is a boot-time denial of service with limited integrity impact rather than confirmed code execution, since the device must be actively NFS-booting and the attacker must control or intercept the NFS server on the boot network. A vendor patch exists (commit 0bbf0985, fixed in 2026.10-rc5) and no public exploit code was identified at time of analysis.
Memory corruption in U-Boot's IP fragment reassembly code (net/net.c) allows remote, unauthenticated attackers on the boot network to crash the bootloader and corrupt adjacent memory while a device is performing a netboot. Only builds compiled with CONFIG_IP_DEFRAG are affected - an opt-in, frequently disabled compile option - so deployments without it are not vulnerable, and exploitation requires delivering a crafted IP fragment with a non-zero offset and the More-Fragments flag onto the boot path during the narrow netboot window. The CVSS 4.0 base score is 8.8 (CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:L/VA:H), an availability-dominant flaw with only limited integrity impact and no confidentiality loss, and no public exploit has been identified at time of analysis (no CISA KEV listing).
Authentication bypass in the optional sshd-ldap component of Apache MINA SSHD (versions 1.2.0 through 2.19.0 and 3.0.0-M1 through 3.0.0-M5) lets unauthenticated remote attackers complete SSH authentication by submitting the literal username "*" with the password "*", because LDAP filter metacharacters from user-supplied credentials are not escaped before being embedded in the search filter. Rated CVSS 9.1 (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N), the flaw is trivially reachable but only on servers that both embed the MINA SSHD library and have deployed and configured sshd-ldap for password or publickey authentication; stock MINA SSHD deployments without that optional component are entirely unaffected, and no public exploit code was identified at time of analysis. Apache ships fixes in 2.20.0 and 3.0.0-M6, which apply proper RFC 4515 escaping.
Authentication bypass in the optional sshd-ldap component of Apache MINA SSHD allows unauthenticated remote attackers (CVSS:3.1/AV:N/AC:L/PR:N/UI:N) to pass password authentication checks when an SSH server is configured to use LdapPasswordAuthenticator, yielding high confidentiality and integrity impact (rated 9.1 critical). Affected builds are 1.2.0 through 2.19.0 and 3.0.0-M1 through 3.0.0-M5; servers that rely on sshd-core's built-in password mechanisms, or that do not use sshd-ldap at all, are not affected. No public exploit code has been identified and the issue is not confirmed actively exploited (no CISA KEV listing), but real-world exposure is sharply limited by the non-default optional component and the explicit LDAP authenticator configuration required - the raw 9.1 score is driven by impact rather than by a broad exploitable population.
Memory exhaustion in the Apache MINA SSHD SFTP client (DefaultSftpClient, sshd-sftp component) lets an attacker-controlled SFTP server drive a client application to run out of heap and crash, because the client stores unsolicited SFTP replies that it never matches to an outstanding request and therefore never releases. The flaw affects client-side deployments of Apache MINA SSHD versions 0.9.0 through 2.19.0 and 3.0.0-M1 through 3.0.0-M5; server-only deployments are not exposed, and the victim must initiate an outbound SFTP session to a malicious or compromised server. It is rated CVSS 3.1 7.5 (AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H) and is a pure denial-of-service condition of the client process, with no authentication bypass, code execution, or data disclosure; no public exploit code was identified at time of analysis and the issue is not listed as confirmed actively exploited (CISA KEV).
A partial authentication bypass in Apache MINA SSHD lets a remote attacker satisfy a repeated-public-key multi-authentication policy by presenting the same public key twice, so a server configured to demand two distinct keys can be entered with only one of them. Affected are versions before 2.20.0 and 3.0.0-M1 through 3.0.0-M5 of the sshd-core component; the vendor assigns CVSS 3.1 8.1 (High, CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:N), where PR:L reflects that the attacker must already legitimately possess at least one accepted key pair. Exploitation is confined to the non-default hardened configuration that requires repeated/multiple public keys (the equivalent of OpenSSH's AuthenticationMethods "publickey,publickey"); single-key or password-based deployments are not affected, and no public exploit code or confirmed active exploitation was identified at time of analysis.
An authentication bypass in Apache MINA SSHD's sshd-core library (versions 2.0.0 through 2.19.0 and 3.0.0-M1 through 3.0.0-M5) allows remote, unauthenticated attackers to skip signature verification during public-key or hostbased SSH authentication, potentially granting an attacker a valid session. The CVSS 3.1 vector AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N (9.1 critical) reflects high confidentiality and integrity impact, but exploitation is gated by a non-default precondition: the server operator must have written explicit custom code enabling the library's 'asynchronous authentication' mechanism and applied it to public-key or hostbased authentication. Apache rates this implementation choice as presumed rare, and servers that do not opt into asynchronous authentication - or that use it only with password or keyboard-interactive auth - are not exposed. No public exploit code has been identified at time of analysis and the vulnerability has not been confirmed as actively exploited (no CISA KEV listing).
Signed integer overflow in the s2 dictionary API of klauspost/compress before 1.18.7 lets an attacker crash a consuming Go process: a dictionary carrying a uvarint-encoded repeat value larger than MaxInt64 slips past repeat-index validation, and a subsequent call to Dict.Encode() converts the overflowed negative repeat value into an out-of-bounds unsafe.Pointer access that aborts the process with SIGSEGV. The issue is availability-only (C:N/I:N, A:H) on a CVSS:3.1/AV:N/AC:L/PR:N/UI:N vector, but real exposure hinges on the application - it only matters where untrusted, attacker-influenced bytes reach s2.NewDict() and the resulting dictionary is then encoded, not in ordinary compression or decompression of arbitrary data. No public exploit code and no CISA KEV entry were identified at time of analysis; the vendor-released patch is v1.18.7.
Heap-based out-of-bounds write in the RPM Package Manager (rpm) affects Red Hat Enterprise Linux 6 through 10 and Red Hat Hardened Images, allowing a local, unauthenticated attacker to corrupt memory by crafting an RPM package that is then parsed with rpm2cpio, rpm2archive, or rpm -qlvp. The flaw arises from an integer overflow in iterReadArchiveNext() when a symlink entry declares RPMTAG_LONGFILESIZES as 0xFFFFFFFFFFFFFFFF, shrinking the buffer allocation to one byte and enabling attacker-controlled data from the cpio filesize field to overwrite heap memory. Exploitation requires user interaction to inspect the malicious package and no public exploit code has been identified at time of analysis; the CVSS 3.1 base score is 7.1.
Local privilege escalation in Ghostscript for Windows (all versions prior to 10.08.0) allows any authenticated local user to plant a malicious PostScript resource file in the predictable C:\gs\ path, which Ghostscript then auto-loads with the privileges of whichever principal next invokes it. The hijack requires no special Ghostscript configuration because it abuses Windows default ACLs that let authenticated users create directories at the root of C:\, and it succeeds only while C:\gs\ is unpopulated - the default state on Windows installations; execution is passive and fires when a user or service subsequently runs Ghostscript, with the payload's privilege gain bounded by that invoking principal (assessed as scope-changed impact). No public exploit code has been identified at time of analysis, and the issue, reported by CERT-PL, is fixed in 10.08.0.
Malformed SCTP traffic can crash the Wireshark/tshark dissector in versions 4.6.0 through 4.6.8 and 4.4.0 through 4.4.18, turning a heap-based buffer overflow (CWE-122) in the SCTP dissector into a denial of service against the analyst's own capture session rather than against any remote service. Exploitation requires the tool to process an attacker-supplied packet - either live-captured on a monitored interface carrying crafted SCTP traffic or opened from a malicious capture file - and the packet must be specifically malformed, which is why the high attack-complexity rating applies. Access requirements are unauthenticated per the CVSS vector (PR:N), and the realistic outcome is a crash of the Wireshark/tshark process, not code execution or host compromise; no public exploit code and no confirmed active exploitation were identified at time of analysis. Note that the originally supplied CVSS vector claimed high confidentiality and integrity impact, which the independent assessment corrected to availability-only (C:N/I:N/A:H).
Wireshark 4.6.0 through 4.6.8 and 4.4.0 through 4.4.18 can be crashed by processing attacker-controlled SPDY protocol data, an availability-only denial-of-service condition that our independent assessment scores as CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:L. The signals conflict and, on balance, point to a LOWER real-world priority than the headline 8.8 suggests: exploitation requires the victim to open a crafted capture file or live-capture crafted SPDY traffic, publicly available exploit code exists, EPSS is only 0.45% (37th percentile), and there is no confirmed active exploitation (CISA KEV). Unauthenticated attackers can trigger the crash, but user interaction is mandatory and the impact is limited to terminating the Wireshark process, not code execution or data compromise.
Path traversal in Wireshark's configuration-profile import handler (CWE-22) allows a maliciously crafted profile to write files outside the intended profile directory, crashing the application and potentially achieving code execution in the user's context. Affected releases are Wireshark 4.4.0 through 4.4.18 and 4.6.0 through 4.6.8; the flaw is rated CVSS 7.8 (AV:L/AC:L/PR:N/UI:R) and requires the victim to manually import the attacker-supplied profile, so exploitation depends on social engineering rather than a network-facing or automatic trigger. Publicly available exploit code exists and the issue is tracked as wnpa-sec-2026-106, but it is not confirmed as actively exploited (CISA KEV) and EPSS is low at 0.19% (7th percentile), making this a moderate-priority issue rather than an emergency; patched releases 4.4.19 and 4.6.9 are available from the vendor.
Use-after-free in libpng 1.6.0 through 1.6.58 can crash applications that use the sequential reader and call png_read_end after png_read_info without first reading the image rows. A crafted PNG containing a zTXt, iTXt or iCCP chunk that makes libpng abandon zlib decompression early leaves a stale input pointer in the zlib stream; the zTXt/iTXt path then reads through a freed heap buffer (a larger tEXt or pCAL chunk before IDAT must first reallocate the read buffer), while the iCCP path is a stack use-after-return. Impact is confined to denial of service - the dangling pointer is only read into a discarded scratch buffer, so there is no data disclosure or corruption - and the required API call sequence is rare in practice; no public exploit has been identified at time of analysis, and the vendor rates it 5.9 Medium (CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H), consistent with the independent assessment.
Signature verification bypass in Authlib 1.7.2 and earlier allows unauthenticated remote attackers to have crafted JWS JSON Serialization payloads accepted as valid, because JsonWebSignature.deserialize_json() returns the payload without checking for a signature or requiring a cryptographic key. The flaw affects applications that use that method to process incoming tokens and then trust the returned payload for authentication, authorization, or integrity decisions; applications using only compact JWS or independently verifying signatures are not exposed. No public exploit code has been identified at time of analysis, and EPSS is low at 0.12% (2nd percentile), but the issue is remotely exploitable with no authentication or user interaction and is assessed as high severity.
Authentication bypass in the LiteLLM proxy lets any party who already holds a valid JWT issued by the configured external identity provider impersonate an existing user - including accounts with proxy_admin privileges - by presenting a token whose unverified email claim matches the victim's address. The JWT authentication flow performs an email-based fallback lookup without validating the email_verified claim, so the attacker inherits the victim's role and can permanently overwrite the victim's stored identity binding, retaining persistent access to administrative endpoints that expose API keys and user management. The attacker must possess a legitimately signed token from the same IdP (CVSS PR:L, AC:H), deployments that do not use JWT/IdP authentication are unaffected, and publicly available exploit code exists; no confirmed in-the-wild exploitation has been reported.
Type confusion in XS::Parse::Infix versions 0.40 through 0.49 - the Perl XS module distributed as part of XS-Parse-Keyword - lets an attacker who can influence a list handed to a list-associative infix operator (for example the zip operator from Syntax::Operator::Zip) make the interpreter dereference an address of the attacker's choosing. Supplying a bare number where an array reference is expected reliably segfaults the process (denial of service), while a crafted string carrying 0x0b at byte 12 slips past the faulty reference guard and is treated as an AV head, so bytes 16-23 are used as AvARRAY and its entries are pushed onto the Perl stack as live SVs, yielding a limited out-of-bounds read / information-disclosure primitive. Publicly available exploit code exists (a minimal zip-based proof-of-concept is described by the reporter), but there is no confirmed active exploitation (no CISA KEV listing) and EPSS is low at 0.18% (7th percentile); this is a library flaw with no standalone attack surface, so real-world risk depends entirely on whether a consuming application passes untrusted, attacker-influenced data into such an operator.
Stored cross-site scripting in the SUSE Rancher UI allows unauthenticated remote attackers to inject malicious script through updates to public UI settings, which then executes in the browser of any user-including administrators-who later views the affected interface. Affected versions are SUSE Rancher 2.15 before 2.15.2, 2.14 before 2.14.6, 2.13 before 2.13.10, 2.12 before 2.12.14, and 2.11 before 2.11.18. Vendor CVSS 3.1 rates this 9.6, while our independent assessment lowers the availability impact to low; no public exploit identified at time of analysis, and exploitation requires network reachability to the Rancher management endpoint plus a victim loading the UI.
Remote attackers holding a previously issued Rancher session token can retain access even after the legitimate user logs out in SUSE Rancher 2.15 before 2.15.2, because the server fails to clear credentials on logout (CWE-613). This session persistence flaw carries a CVSS 3.1 base score of 8.1 (High) with network reach, low complexity, no attacker privileges (PR:N), and required user interaction (UI:R) for the logout. No public exploit identified at time of analysis; exploitation is limited to scenarios where the attacker already possesses a valid token, such as shared-device or stolen-session situations.
Privilege escalation in SUSE Rancher Fleet (0.14.x before 0.14.11, 0.15.x before 0.15.7, and 0.16.x before 0.16.2) lets a low-privileged tenant - an authenticated user able to push a Fleet Bundle or GitRepo - have resources written to a shared downstream cluster under the Fleet agent's own cluster-admin credentials rather than the ServiceAccount pinned to the deployment, allowing that tenant to overwrite configuration owned by other tenants. Exploitation is gated by topology: it only matters in multi-tenant Fleet deployments where privileged and untrusted teams share the same downstream cluster, since single-tenant or per-tenant-isolated clusters have no crossed trust boundary. Rated 8.8 by the vendor (the independent assessment scores it higher, 9.9, because the trust boundary crossing changes scope); no confirmed active exploitation (CISA KEV) and no public exploit identified at time of analysis.
Path traversal in Flatpak's app deployment handling of the exported files/etc directory allows a malicious Flatpak application to empty or replace host files such as /etc/passwd, /etc/group, /etc/machine-id, or /etc/resolv.conf with a symlink when that app is installed or upgraded; on system-wide installations the write is carried out as root. Exploitation is gated on user interaction: the victim must choose to install or upgrade the attacker-supplied app from an untrusted source (CVSS 3.1 base 7.1, AV:N/AC:L/PR:N/UI:R with high availability and low integrity impact), making this a genuine but moderate-priority integrity/availability issue rather than a critical remote code execution flaw. No public exploit code and no confirmed active exploitation (CISA KEV) were identified at time of analysis.
Path traversal in Flatpak's handling of the export/bin directory during application deployment allows a malicious Flatpak app to delete attacker-chosen files outside the deployment directory when that app is installed or upgraded; on system-wide installations the deletions are carried out with root privileges, so a crafted package can remove arbitrary files anywhere the deployer can reach. The flaw affects Flatpak releases before 1.18.4 and is triggered by a user or administrator installing or upgrading an attacker-supplied app, which supplies the required user interaction rather than a network-based trigger; the CVSS vector carries PR:N, so no prior authentication is needed beyond that install action. No public exploit code has been identified at time of analysis and there is no evidence of confirmed active exploitation, though the integrity and availability impact is high (arbitrary file deletion) and the flaw crosses a privilege boundary on system-wide installs.
Cross-tenant authorization bypass in SUSE Rancher Fleet allows a party who can perform agent-initiated cluster registration into a Fleet workspace namespace shared with other tenants to inject attacker-controlled cluster labels - including reserved management.cattle.io/ labels such as the cluster display name - that are then persisted onto the upstream Cluster object. Because Fleet resolves GitRepo and Bundle targeting from those labels, the attacker's own cluster can be made to satisfy targeting rules administrators intended for a different tenant's cluster, causing content such as Git repos and bundles to be deployed to the wrong cluster. The flaw affects Fleet 0.16 before 0.16.1, 0.15 before 0.15.6, 0.14 before 0.14.10, 0.13 before 0.13.15, 0.12 before 0.12.19 and older releases; exploitation is conditional on a multi-tenant deployment with a shared workspace namespace and label-based targeting, requires a low-privileged authenticated position (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:N, 7.1), and no public exploit code has been identified at time of analysis.
Path traversal in the btrfs storage driver of Canonical LXD 4.0.2 and later allows an authenticated client that holds project-scoped permission to create instances to delete arbitrary files on the host as root by supplying a crafted subvolume path containing ../ sequences, and on hosts whose root filesystem is itself btrfs the same primitive escalates to placing attacker-controlled content at arbitrary host paths for full host compromise. The issue is rated CVSS v3.1 9.6 (AV:N/AC:L/PR:L/UI:N/S:C/C:N/I:H/A:H), reflecting a scope-changing breakout from a restricted tenant role into the host root context rather than a CVSS-inflated score, since a legitimate but limited project credential is the only prerequisite. No CISA KEV listing and no public exploit code were identified at time of analysis; vendor fixes are available in LXD 4.0.14, 5.0.10, 5.21.8 and 6.10.
Path traversal in Canonical LXD's Btrfs storage driver allows an authenticated user with instance-creation privileges to escape the intended backup import target and delete or replace arbitrary files and directories on the host as root, by supplying a crafted subvolumes[].path value in backup/optimized_header.yaml during a btrfs 'optimized' backup import. The assessed vector (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H, 9.9) reflects high confidentiality, integrity and availability impact with a scope change because the traversed write lands outside the LXD daemon's own container boundary; exploitation requires valid low-privilege API credentials and a Btrfs-backed host, not anonymous access. No public exploit code has been identified at time of analysis, and the issue is not confirmed actively exploited (CISA KEV), but the low attack complexity and root-level file-system effect make it a genuine high-priority fix rather than a CVSS-inflated score.
OS command injection in SUSE NeuVector 5.4 (before 5.4.11), 5.5 (before 5.5.4), and 5.6 (before 5.6.2) allows an attacker who already satisfies one of two prerequisites - an authenticated NeuVector user holding the namespaced Runtime Policies (write) permission, or possession of NeuVector's internal gRPC certificate key pair - to inject operating-system commands that execute inside the privileged enforcer container, resulting in complete compromise of the underlying worker node. The flaw is rated 8.8 (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H), meaning remote reachability, low attack complexity, authenticated low privileges, and no user interaction, with total confidentiality, integrity, and availability impact once the prerequisite is met. Ordinary read-only users and unauthenticated network clients cannot trigger it, but any deployment that grants Runtime Policies write access broadly or leaks the internal gRPC key pair is at high risk; no public exploit code has been identified at time of analysis and the issue is not listed in CISA KEV, though weaponization would be straightforward given the trivially reachable command sink.
An out-of-bounds read in libXtst's RECORD reply parser (CWE-126) allows a malicious or compromised X server to crash attached X clients that link the library and use the XTEST/RECORD extension, affecting all versions before 1.2.6. The assessed vector is CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:L/I:N/A:L, meaning unauthenticated attackers in the network position of the rogue server can trigger a low-integrity information leak and a denial of service, but only if the victim voluntarily connects to that server (UI:R). This is a reverse-trust scenario: it is not exploitable against clients that connect only to trusted local or remote X servers, and the X server itself is unaffected. There is no public exploit identified at time of analysis and the issue is not in CISA KEV; vendor patch data indicates a fix is available, with the upstream correction landing in the libXtst 1.2.6 line.
A race condition in the Linux kernel cgroup task iterator (css_task_iter_next) can hand back a task_struct whose reference count has already dropped to zero, resulting in a use-after-free that a local attacker can turn into kernel memory corruption. Only systems running affected kernels built with cgroups are exposed, and the attacker must be able to read a cgroup.procs file while concurrently spawning and tearing down multi-threaded process groups - a low-privileged but purely local position (CVSS:3.1/AV:L/AC:H/PR:L). Exploitation is gated by an extremely narrow timing window tied to RCU grace periods and task teardown ordering, so the flaw is high-severity on paper but difficult to trigger reliably; EPSS is only 0.18% (7th percentile) and no public exploit code was identified at time of analysis.
Path traversal via symlink prefix reuse in gitoxide's gix-fs crate (before 0.23.0; GHSA lists <= 0.21.0) allows a crafted repository tree to write attacker-controlled symlinks outside the worktree during forced checkout with overwrite_existing enabled, enabling code execution by overwriting known paths such as .git/hooks/post-checkout. Exploitation is local, requires user interaction to initiate checkout of an untrusted tree, and has high attack complexity tied to winning the prefix-reuse condition; no public exploit was identified at time of analysis, and no KEV or EPSS data were provided. Per the assessed CVSS:3.1 vector AV:L/AC:H/PR:N/UI:R/S:U/C:H/I:H/A:H, unauthenticated but local access with user interaction is required.
Reflected cross-site scripting in the RabbitMQ management UI lets an attacker execute arbitrary JavaScript in the management interface origin by injecting content into a bearer token that is embedded unescaped in the OAuth bootstrap script generated by set_token_auth/2. RabbitMQ server 4.2.0 through 4.2.7 and 4.3.0/4.3.1 are affected only when the management plugin is configured with management.oauth_enabled=true, which is what exposes the OAuth bootstrap endpoint before authentication; the attacker must also induce an administrator to load the affected page (UI:R), and the cookie-based variant additionally requires the attacker to first plant an access_token cookie on the management host. The vendor-supplied CVSS 4.0 base score of 7.5 is higher than our independent CVSS 3.1 assessment (AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:L/A:N), which better reflects the non-default configuration gate and the extra cookie-planting prerequisite; no public exploit code has been identified at time of analysis.
A tenant isolation bypass in BerriAI LiteLLM versions before 1.101.0-rc.1 lets an authenticated holder of a valid virtual key read other tenants' entries from the semantic cache and, in the worst case, have attacker-supplied tool calls executed under a victim's credentials. The flaw stems from an inconsistency in how the cache layer derives the tenant scope versus the metadata variable name, so semantically similar prompts sent to routes such as /v1/responses and /bedrock/* can collide with and return another tenant's cached response - potentially exposing PII, financial data, or source code. Exploitation requires a multi-tenant deployment with semantic caching enabled and a sufficiently similar prompt to hit the victim's cache entry, and it is rated CVSS 8.7 with a vendor patch already released (commit 16db51e2, fixed in v1.101.0-rc.1); no public exploit code has been identified at time of analysis and the CVE is not in CISA KEV.
A use-after-free race condition in QEMU's 9pfs (virtfs) passthrough subsystem allows a malicious guest user to escape the shared directory boundary and read or write arbitrary files on the host, potentially achieving code execution as the QEMU process user (VM escape). Only hosts that have enabled 9pfs/virtfs host-directory sharing are exposed, since this is not a default configuration; the attacker must be a local guest user able to issue 9p protocol requests and must win a race between QEMU's main thread and a 9pfs worker thread with concurrent Tlcreate and Twalk requests. No public exploit code or active exploitation telemetry was identified at time of analysis, but the flaw is rated high severity (CVSS 8.8, scope-changed, C:H/I:H/A:H) and affects QEMU packages shipped in Red Hat Enterprise Linux 6 through 10, Red Hat OpenShift Container Platform 4, and related Red Hat products.
Command injection in the kitty terminal emulator (versions 0.47.3 through 0.48.x) lets any program that can get output rendered on screen - a file shown with cat, a log line, command output, an SSH banner, or a web response - make kitty write attacker-composed printable ASCII back into the pseudoterminal, where the user's shell treats it as typed input and executes it with the user's privileges. The root cause is that the OSC 21 colour-control handler echoed unrecognised field names verbatim into its reply, and the 0.47.3 restriction (introduced as the fix for CVE-2026-54057) to printable ASCII excluding ';' and '=' still left enough characters to compose a shell command; the newline needed to submit that command is obtainable through the ssh kitten path. Exploitation is local (AV:L) and requires the victim to display attacker-controlled bytes (UI:P), no privileges are required, and no public exploit code has been identified in the available intelligence at time of analysis. The 9.3 CVSS 4.0 base score overstates real-world reach, since this is not a remote, unauthenticated, one-click compromise; the vendor-released patch is kitty 0.49.0.
The Linux kernel's nvme-rdma host driver mishandles command teardown when a queue_rq submission returns -EIO, unmapping SQE DMA and cleaning up the block-layer request after the host-path-error helper has already completed that request. The result is a double cleanup of the nvme-rdma command and blk-mq request plus a DMA unmap against an already-completed request, which can corrupt request/command state, trigger a kernel crash or denial of service, and - per the assessed vector - cause limited integrity impact. Only hosts running the nvme-rdma driver while connected to an NVMe-oF target over an RDMA fabric (RoCE, iWARP or InfiniBand) are exposed, and the faulting branch only executes on a transient RDMA command-submission failure that is not directly attacker-controllable; the local, high-complexity, privileged nature of the trigger makes real-world risk substantially lower than the base score suggests, and no public exploit identified at time of analysis.
Improper CPU-ID validation in the Linux kernel's BPF subsystem (the BPF_F_CPU map-operation flag) allows a local user with BPF privileges - CAP_BPF, or any environment where unprivileged BPF is enabled - to drive an out-of-bounds per-CPU map access on hosts whose possible-CPU mask is sparse/non-contiguous, producing a kernel paging fault and denial of service (assessed availability impact A:H, with only limited integrity impact and no confidentiality impact). Only kernels exposing the relatively new BPF_F_CPU flag (introduced around Linux 7.0) are in scope, and exploitation additionally requires a machine with a CPU ID gap such as 0,2-3 - typical of certain arm64/embedded boards or virtualized guests with CPU device-tree holes - while mainstream x86 systems with contiguous 0..N-1 CPUs are not affected because num_possible_cpus() equals nr_cpu_ids there. There is no remote vector, exploitation is local and requires BPF privileges (CVSS vector PR:L, AC:H), and no public exploit code was identified at time of analysis; no CISA KEV entry was provided, so active exploitation is not confirmed, and this is best characterized as a high-CVSS-but-low-real-risk issue given the sparse-mask and CAP_BPF prerequisites.
Incorrect interpretation of the NPU_SET_IFM/OFM_PRECISION field by the Linux kernel's accel/ethosu driver on Arm Ethos-U65 hardware lets a local user with access to the accelerator character device corrupt command-stream processing. Because bits 15:14 are read as an activation storage mode instead of the U65 rounding mode, natural rounding (value 2) streams are rejected with -EINVAL, and truncating rounding (value 1) streams are misrouted to the two-tile address path where a feature-map size is computed from tile bases the stream never initialized, degrading integrity of NPU operations and potentially crashing the affected subsystem. Authentication is required (CVSS vector PR:L, local vector AV:L); no public exploit code has been identified at time of analysis, and exploitation is limited to systems running the driver bound to real Ethos-U65 silicon, such as NXP i.MX93 SoCs. Vendor-released patches are available in the referenced kernel stable commits, with EUVD listing Linux 7.2.7 and 7.3-rc2 as fixed.
Linux kernel SCTP timer handling is vulnerable to a time-of-check-to-time-of-use race that can unbalance an association's reference count and cause a use-after-free during association teardown, allowing an unauthenticated remote peer to crash or destabilize an affected system. Any kernel built with the SCTP protocol enabled (CONFIG_IP_SCTP) and exposing an SCTP endpoint to untrusted peers is in scope; the assessed vector is AV:N/AC:H/PR:N/UI:N with no confidentiality impact, low integrity impact, and high availability impact, because winning the narrow timing race is the attacker's main obstacle. There is no public exploit identified at time of analysis and no CISA KEV listing; the upstream and stable-branch fixes are already published, and no EPSS score was supplied in the available intelligence.
Use-after-free in the Linux kernel's VXLAN multicast database (MDB) handling lets a local, low-privileged attacker corrupt kernel memory by crafting netlink MDB requests, though no public exploit code was identified at time of analysis and the issue is not confirmed as actively exploited (CISA KEV). The root cause is that vxlan_mdb_is_valid_source() accepts the all-zeros address as both MDBE_ATTR_SOURCE and a source-list member, causing the derived (S,G) key to be byte-identical to the (*,G) key, so vxlan_mdb_remote_src_del() frees an entry and then deletes it a second time. The independently assessed impact is limited to integrity and availability (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:L/A:H) rather than the higher confidentiality/integrity scores in the vendor-published vector, because the flaw is a local write primitive and denial-of-service/crash risk rather than a disclosure channel; exploitation requires a user namespace where an unprivileged user can obtain CAP_NET_ADMIN in a new network namespace, and is blocked where unprivileged user namespaces are disabled.
A page-level use-after-free in the Linux kernel ALSA PCM subsystem lets a local user with access to a /dev/snd PCM character device turn a race between a concurrent mmap() and a buffer reallocation into a stale, writable mapping of freed DMA buffer pages, which can be reclaimed by the allocator and reused for local privilege escalation. The attacker must issue SNDRV_PCM_IOCTL_HW_PARAMS/HW_FREE from one thread while another thread mmaps the same PCM stream, and must win a narrow timing window (assessed CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H); no public exploit code has been identified at time of analysis, although the upstream commit states a reproducer is available on request, and this CVE is not in CISA KEV. Vendor patches are available in the stable trees (6.12.111, 6.18.53, 7.2.7) and mainline 7.3-rc2, with backports to older long-term branches, so unpatched kernels running ALSA PCM support remain exposed to any local user who can open a PCM device.
A use-after-free race condition in the Linux kernel's ksmbd in-kernel SMB3 server can crash the kernel (denial of service) when a session is torn down while multiple channels are bound to a session and deferred byte-range locks or pending CANCELs remain counted as running requests. Affected systems must have ksmbd enabled (CONFIG_SMB_SERVER) and exporting at least one SMB3 share, and an authenticated remote attacker (PR:L) must negotiate SMB3 multichannel to race a LOGOFF, synchronous CANCEL, or abrupt connection teardown against still-running requests; success is non-deterministic (AC:H). No public exploit code has been identified at time of analysis, and the impact is availability loss, not information disclosure or confidentiality/integrity compromise as some tags or the base CVSS may imply.
Use-after-free race in the Linux kernel's ksmbd SMB3 server can be triggered when a network-device event such as an interface up/down operation invokes ksmbd's netdevice notifier during listener teardown, potentially allowing a local actor who can induce netdev events to corrupt memory and cause denial of service. Affected systems are those with CONFIG_SMB_SERVER enabled and at least one ksmbd listener bound to a network interface; hosts using userspace Samba instead of ksmbd are not affected. No CISA KEV entry or public exploit code was provided, and no public exploit identified at time of analysis; the assessed vector is local, high complexity, PR:L, C:N/I:L/A:H, so the NVD 7.8 vector overstates confidentiality and integrity risk, while vendor-released fixes are available in Linux 7.2.7 and 7.3-rc2.
A stack out-of-bounds write in the Linux kernel's Bluetooth L2CAP LE connection path lets an attacker within Bluetooth LE radio range corrupt kernel memory by sending L2CAP_LE_CONN_REQ frames - including duplicate identifiers - to a host running a userspace Bluetooth application that listens on an L2CAP LE channel with L2CAP_MODE_EXT_FLOWCTL and FLAG_DEFER_SETUP set. The CVSS base score is 7.5, but the independent assessment rates actual impact as integrity and availability only (no confidentiality impact), gated on that specific, non-default socket configuration, physical radio proximity, and affected kernels from 5.7 onward; no public exploit code has been identified and no active exploitation is confirmed. Unauthenticated (PR:N) adjacent attackers are the relevant threat model per the CVSS vector, and vendor patches are available.
Out-of-bounds memory access in the Linux kernel's IPv6 Segment Routing (SRv6) receive path can be triggered by remote, unauthenticated attackers who send a crafted IPv6 packet containing a Segment Routing Header (SRH) with segments_left > 0 that is preceded by another extension header such as a Hop-by-Hop options header. Because ipv6_srh_rcv() restores skb->data by pushing a fixed sizeof(struct ipv6hdr) instead of re-establishing the real network header, skb_network_offset() remains negative; that negative value is then consumed as an unsigned length by the flow dissector during ip6_route_input(), or by sch_fragment()/neighbour output when the packet is forwarded via ip6_forward() or redirected via act_mirred, producing an out-of-bounds memcpy or buffer overflow that can crash the kernel and corrupt memory. Exploitation is not automatic: the target must be built with CONFIG_IPV6_SEG6 and configured as an SRv6 endpoint, and the packet must be specifically structured, which is reflected in the assessed vector AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:H; no public exploit code and no confirmed active exploitation were identified at time of analysis. Vendor fixes are available in Linux 6.12.111, 6.18.53, 7.2.7 and 7.3-rc2.
A use-after-free in the Linux kernel's btrfs RAID stripe-tree insertion path can free an in-flight transaction handle and transaction twice, corrupting kernel memory on systems that enable the opt-in raid-stripe-tree feature (added in Linux 6.7). Only local users who can write to such a filesystem, and only when a memory allocation fails (-ENOMEM) at exactly the btrfs_insert_one_raid_extent() failure site during ordered-extent finishing, can trigger it; the independent assessment scores this AV:L/AC:H/PR:L with no confidentiality impact but high availability impact (C:N/I:L/A:H), noting it is a high-CVSS-but-low-real-risk kernel issue that is not attacker-input-controlled. There is no public exploit identified at time of analysis and the flaw is not confirmed actively exploited (CISA KEV); vendor-released patches exist in the affected branches.
A local altname-conflict handling flaw in the Linux kernel's network namespace teardown path (__dev_change_net_namespace()) allows an attacker with CAP_NET_ADMIN in a network namespace to trigger a WARN_ON assertion splat in cfg80211_pernet_exit() by assigning the same alternative interface name to devices in both the root netns and a non-root netns and then destroying the non-root netns. The affected code is the wireless (cfg80211/cfg802154) per-netns cleanup, which previously lacked the conflict-removal logic already added for default_device_exit_net(); the fix generalizes that cleanup into __dev_change_net_namespace(). Exploitation requires local access and the ability to create network namespaces and devices - achievable by an unprivileged user where unprivileged user namespaces are enabled - but the practical impact is limited to a kernel warning on default configurations, escalating to a real denial of service only when panic_on_warn=1 is set; no public exploit code has been identified at time of analysis, and this issue is not confirmed actively exploited (CISA KEV). Our independent assessment rates this as low real risk and considers the supplied CVSS 7.8 (C:H/I:H/A:H) materially overstated.
A race condition in the Linux kernel's RDS-over-TCP transport (net/rds) lets a local user who can open and use an AF_RDS socket corrupt connection state when a transmit is in flight at the same moment the accept path resets the connection's callbacks. Because rds_tcp_reset_callbacks() only samples the RDS_IN_XMIT bit as clear rather than owning it, rds_send_xmit() can re-take the bit immediately afterwards and run concurrently with rds_send_path_reset() rewriting cp_xmit_* state, producing torn transmit state and a potential kernel crash/denial of service with limited integrity impact. The authoritative assessment scores this CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:H - local access with low privileges, a narrow timing window, and, critically, manifestation restricted to weakly ordered CPUs (ARM, PowerPC) since the store-buffering pattern cannot occur on strongly ordered x86; no public exploit code has been identified at time of analysis and there is no confirmation of active exploitation (no CISA KEV entry), although the public tag of 'Information Disclosure' is a mischaracterization of what is really a CWE-362 concurrency bug. Vendor-released patches exist for the stable branches (6.12.111, 6.18.53, 7.2.7, and 7.3-rc2).
The Linux kernel's Reliable Datagram Sockets (RDS) connection teardown path contains a race condition in rds_conn_shutdown() that can crash or hang the kernel, causing denial of service on systems where the RDS subsystem is loaded and connections are being shut down or failed over. The flaw stems from teardown merely sampling the RDS_IN_XMIT and RDS_RECV_REFILL fastpath bits as clear instead of owning them, so a concurrent rds_send_xmit() or rds_ib_recv_refill() can re-acquire the bit and run while the transport re-initializes its rings and rds_send_path_reset() rewrites transmit state under it. Because the store-buffering pattern only manifests where memory ordering is weak (for example ARM), and because the accept-side state check must pass before the path moves to RDS_CONN_DISCONNECTING, reliable triggering is difficult and the practical impact is limited to kernel instability; no public exploit code has been identified at time of analysis and vendor patches are available.
A soft-lockup denial of service in the Linux kernel NVMe driver lets an attacker-controlled or misbehaving NVMe target stall host CPUs by advertising a sparse 32-bit namespace ID list containing a large gap, because nvme_scan_ns_list() walks every missing NSID one at a time instead of only the namespaces actually present. Practical exploitation is not drive-by: the host administrator must first attach the host to the hostile target, either over NVMe-over-Fabrics (an 'nvme connect' style operation) or through a crafted local NVMe device, and the resulting impact is availability only (C:N/I:N, A:H; assessed vector CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H, feed score 7.5). No public exploit identified at time of analysis and no CISA KEV entry is associated with this CVE, but vendor-released patches exist in the 6.12.111, 6.18.53, 7.2.7 and 7.3-rc2 kernel releases.
Use-after-free in the Linux kernel's bcmasp driver for Broadcom ASP 2.0 Ethernet hardware can cause premature release of an sk_buff when a reused TX descriptor retains a stale "last" flag, affecting only systems with that specific Broadcom NIC/SoC and requiring local access plus fragmented transmission activity that leads to descriptor reuse. The assessed impact is primarily availability (A:H) with limited integrity (I:L) and no confidentiality loss (C:N) under CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:H, and no public exploit identified at time of analysis. The original base score of 7.8 and C:H/I:H/A:H overstate real-world risk because the trigger requires specific hardware, local privileges, and a non-deterministic timing-dependent TX-ring condition.
A concurrency flaw in the Linux kernel's mlxsw Spectrum PTP driver (spectrum_ptp) can panic the kernel when a stale PTP entry is reaped by the garbage-collection workqueue at the same moment the threaded NAPI poll is running on another CPU. The GC path calls mlxsw_sp1_ptp_unmatched_finish() → mlxsw_sp1_ptp_packet_finish() → mlxsw_sp_rx_listener_no_mark_func(), which ends in napi_gro_receive() using a NAPI pointer captured earlier in the RX context; because mlxsw_pci_napi_devs_init() unconditionally calls netif_threaded_enable(), that poll runs as an independently scheduled kthread that local_bh_disable() on the GC CPU cannot protect against, so both CPUs mutate the same GRO list and hit list_add corruption / 'kernel BUG at lib/list_debug.c'. Exploitation is not remotely trivial: it requires physical Mellanox Spectrum hardware running the mlxsw driver, PTP hardware timestamping configured and active (e.g. ptp4l) with privileged CAP_NET_ADMIN access, plus a genuine timing race (AC:H) - the reported trigger is a tx-timestamp timeout while a port transitions to UNCALIBRATED/SLAVE. No public exploit code or in-the-wild activity has been identified at time of analysis, and our independent assessment rates the published CVSS 7.5 (AV:A/PR:N) as an overstatement, scoring the issue as a locally triggered, privilege-gated denial-of-service/kernel-panic (CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:H).
Local privilege escalation in the kernel-mode layer of the NVIDIA GPU Display Driver for Linux allows an unprivileged local user to bypass read-only memory protection and gain write access to memory the driver marked read-only, which can result in kernel code execution, denial of service, privilege escalation, information disclosure, and data tampering. Exploitation requires the attacker to already hold low-privileged local access on an affected Linux host (CVSS 3.1 vector AV:L/AC:L/PR:L/UI:N, score 7.8) with no user interaction and no network vector; the risk is greatest on shared, multi-tenant vGPU, VDI, and cloud-GPU hosts where crossing the user-to-kernel boundary breaks tenant isolation. No public exploit code has been identified at time of analysis and the issue is not listed in CISA KEV, but NVIDIA has published a fix through security bulletin 5861. This is a genuine local privilege-escalation risk rather than an emergency, and it should be prioritized on multi-tenant GPU infrastructure.
Local privilege escalation in NVIDIA Display Driver for Windows and Linux allows an unprivileged local attacker to trigger a use-after-free condition by issuing a crafted sequence of driver commands, potentially achieving code execution, privilege escalation, denial of service, or information disclosure. The vulnerability affects systems with GeForce, RTX/Quadro/NVS, or Tesla GPUs running vulnerable driver versions. No public exploit identified at time of analysis.
Local privilege escalation and memory corruption in NVIDIA GPU Display Driver on Windows and Linux, affecting GeForce, RTX/Quadro/NVS, and Tesla lines. An unprivileged local user can trigger a use-after-free (CWE-416) by issuing a specific sequence of driver commands, leading to potential code execution, privilege escalation, denial of service, or information disclosure. Requires local access and low privileges (AV:L/PR:L) with no user interaction (UI:N); no public exploit identified at time of analysis. Risk is highest in multi-user or shared-GPU environments such as VDI, terminal servers, and cloud/virtualized GPU hosts.
Local privilege escalation in the NVIDIA GPU Display Driver for Linux allows an authenticated, unprivileged user to trigger a use-after-free condition by issuing a specific sequence of driver commands, potentially leading to code execution, privilege escalation, denial of service, and information disclosure. The vulnerability affects GeForce, RTX/Quadro/NVS, and Tesla GPU drivers, including guest-VM vGPU deployments, and requires local access with low privileges. No public exploit code or confirmed active exploitation has been identified at time of analysis.
Use-after-free in the NVIDIA GPU Display Driver for Linux allows a local, unprivileged user to corrupt driver memory by racing or abusing object lifetime handling, with a successful exploit potentially yielding code execution in kernel context, privilege escalation to root, denial of service, information disclosure, and data tampering. The flaw scores 7.8 (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) and is confined to local attackers who already hold an account on the GPU host, with no network vector and no user interaction required. This is a genuine but not emergency-grade issue: no public exploit code has been identified at time of analysis and the CVE is not listed in CISA KEV, so realistic exposure is concentrated on shared or multi-tenant GPU systems - VDI desktops, cloud GPU instances, and HPC nodes - where untrusted users share a single driver instance.
Local code execution and denial of service on Windows systems with affected NVIDIA GPU Display Drivers stem from an integer overflow in the display driver's DIAG escape handler that results in an out-of-bounds write (CWE-787). An unprivileged local user (CVSS PR:L) can trigger the flaw via a device I/O call to the handler; no user interaction is required, and exploitation is confined to the local machine, meaning an attacker must already have a foothold. No public exploit identified at time of analysis (and no CISA KEV entry), but the shared-system impact-multi-user workstations, VDI, and cloud GPU instances-raises the risk profile.
Type confusion in the kernel-mode layer of the NVIDIA GPU Display Driver for Linux allows a local low-privileged user (CVSS vector AV:L/AC:L/PR:L/UI:N) who can reach the driver's device-node/ioctl interface to trigger memory corruption with high impact across confidentiality, integrity, and availability, potentially yielding code execution in kernel context, privilege escalation to root, denial of service, information disclosure, and data tampering. The CVSS 3.1 base score of 7.8 reflects a local attack vector paired with uniformly high impacts rather than any remote reachability, and despite the 'RCE' tag in the source metadata the vulnerability is not remotely exploitable. No public exploit code or confirmed in-the-wild exploitation was identified at time of analysis, and no EPSS or CISA KEV signal was provided.
A type-confusion flaw in the kernel-mode layer of the NVIDIA GPU Display Driver for Linux lets a local, low-privileged user who can open the driver's device nodes and issue ioctls trigger memory corruption by winning a handle-recycle race, yielding kernel-level code execution, privilege escalation, denial of service, information disclosure and data tampering. Affected platforms are the Linux driver branches covering GeForce, RTX/Quadro/NVS, Tesla/data-center GPUs and the vGPU guest driver and Virtual GPU Manager (CVSS 3.1 base 7.8; our independent assessment rates it AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H, i.e. 7.0, because the race window makes exploitation timing-sensitive rather than deterministic). This is a local privilege-escalation-class issue with no remote vector: no public exploit code was identified at time of analysis and the flaw is not confirmed actively exploited (CISA KEV), though it is most consequential on multi-user, VDI and cloud/virtualized GPU hosts where untrusted tenants share a single driver instance.
Local privilege escalation in NVIDIA GPU Display Driver for Linux allows an unprivileged local user to trigger an out-of-bounds write in kernel memory via a crafted ioctl call with an unvalidated buffer size, enabling code execution, privilege escalation, information disclosure, and denial of service. The flaw affects GeForce, RTX/Quadro/NVS, and Tesla drivers, as well as Guest Driver and Virtual GPU Manager components, with highest risk in multi-tenant GPU environments such as VDI and cloud/HPC deployments. No public exploit identified at time of analysis; CVSS 3.1 score is 7.8 (local, low privileges, no user interaction).
Local privilege escalation in the NVIDIA GPU Display Driver for Linux allows an authenticated unprivileged user to trigger a use-after-free in the kernel mode layer, potentially achieving code execution, denial of service, privilege escalation, information disclosure, and data tampering. The flaw is reachable only by a locally authenticated low-privileged account (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) that can issue ioctls to the GPU kernel-mode device interface; there is no remote attack path and no user interaction is required once a foothold exists. The 'Guest Driver' classification makes the issue most consequential in virtualized/vGPU and multi-tenant datacenter environments, where a low-privileged guest can reach the shared GPU driver. No public exploit identified at time of analysis, and the vendor advisory is the authoritative source for fixed driver versions.
Local privilege escalation in NVIDIA's GPU Display Driver (GeForce, RTX/Quadro/NVS, and Tesla, on both Windows and Linux) allows an authenticated low-privileged local user to reach GPU memory belonging to another process, enabling code execution, escalation of privileges, data tampering, denial of service, and information disclosure. Per the assessed vector (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) there is no remote or network path; the flaw is most consequential on shared, multi-user, VDI, or multi-tenant GPU hosts where cross-process GPU memory isolation is the boundary between users and tenants. No public exploit code and no confirmed active exploitation were identified at time of analysis, and the provided vendor reference points to NVIDIA's product security advisory (2026/5861) rather than a fixed driver version named in this data set.
Local privilege escalation in the NVIDIA GPU Display Driver for Linux kernel mode layer allows an authenticated, unprivileged user to trigger a double-free of imported memory state, potentially leading to code execution, denial of service, information disclosure, and data tampering. The flaw requires local access and a low-privilege account; no public exploit code or active exploitation has been identified at time of analysis. The CVSS 3.1 score is 7.8 (HIGH).
Local privilege escalation, code execution, denial of service, and information disclosure in NVIDIA GPU Display Driver on Windows and Linux stem from an integer overflow (CWE-190) in the kernel-mode layer that an unprivileged user can trigger to cause an out-of-bounds write. Exploitation requires local access with an existing low-privilege account (authenticated, PR:L) on a host with an affected GeForce, RTX/Quadro/NVS, or Tesla driver; the CVSS v3.1 base score is 7.8 (High) with a local attack vector (AV:L/AC:L/PR:L/UI:N). No public exploit code or CISA KEV entry was identified at the time of analysis, but the flaw is genuine and actionable for local privilege escalation if the vulnerable driver interface is reachable.
An out-of-bounds write in the kernel-mode layer of the NVIDIA GPU Display Driver for Windows and Linux allows an unprivileged local user to corrupt kernel memory, which can lead to code execution in the kernel, privilege escalation, denial of service, information disclosure, or data tampering. Systems running GeForce, RTX/Quadro/NVS, or Tesla driver packages on either operating system are affected; per the CVSS vector (AV:L, PR:L, UI:N) the flaw is reachable only by a user who already has local code execution on the host, making it a local privilege-escalation primitive rather than a remote entry point. No public exploit has been identified at time of analysis.
Local privilege escalation in NVIDIA GPU Display Driver for Linux stems from a missing authorization check in the kernel mode layer, allowing an unprivileged local user to modify privileged configuration and potentially achieve code execution, escalation of privileges, denial of service, information disclosure, and data tampering. Exploitation requires a pre-existing local low-privileged foothold (CVSS:3.1/AV:L/AC:L/PR:L/UI:N), such as an unprivileged account on a host or inside a guest VM using the NVIDIA guest driver or Virtual GPU Manager, so this is an escalation primitive rather than a remote initial-access vector. No public exploit code is identified and no CISA KEV active exploitation is confirmed at time of analysis, but the high confidentiality, integrity, and availability impacts and broad affected driver families make patching a priority.
Use-after-free in the NVIDIA GPU Display Driver kernel mode layer (Windows and Linux) lets a local user who already holds low-privileged code execution on the host trigger a memory-corruption condition inside the kernel driver, with the potential for code execution, privilege escalation, information disclosure, data tampering, and denial of service. Affected platforms span consumer and professional GPU lines (GeForce, RTX/Quadro/NVS, Tesla) as well as the vGPU components, where a malicious or compromised guest can exercise the vulnerable path through the Guest Driver and Virtual GPU Manager. The attack is strictly local (CVSS:3.1/AV:L/AC:L/PR:L/UI:N, score 7.8), requires no user interaction and no non-default configuration beyond having the driver installed, and there is no public exploit identified at time of analysis - no KEV entry and no proof-of-concept code is referenced in the available data.
A locally-triggered out-of-bounds write in the NVIDIA GPU Display Driver kernel-mode layer lets an attacker who already holds low-level privileges on a Windows or Linux host corrupt kernel memory, with the potential for code execution in kernel context, escalation of privileges, denial of service, information disclosure and data tampering. The affected driver ships by default across the GeForce, RTX/Quadro/NVS, Tesla and Virtual GPU Manager product lines, so any host with a current NVIDIA driver installed is in scope, but exploitation is not remote: an attacker must first obtain code execution on the target (an authenticated local user, a compromised low-privileged process, or sandboxed code with driver access) and then develop a reliable kernel corruption exploit. CVE-2026-47548 carries a vendor/NVD CVSS 3.1 vector of AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8, High); no public exploit code has been identified at time of analysis and the CVE is not listed in CISA KEV.
Local attackers with low-privileged code execution can trigger an out-of-bounds read (CWE-125) in the kernel-mode layer of the NVIDIA GPU Display Driver on Windows and Linux, which may lead to escalation of privileges, code execution, denial of service, information disclosure, and data tampering. This locally exploitable flaw (CVSS 7.8, AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) affects GeForce, RTX/Quadro/NVS, Tesla, and Virtual GPU Manager products, and no public exploit has been identified at the time of analysis. A vendor advisory with remediation guidance is available from NVIDIA via its product security page.
Integer underflow in the kernel-mode layer of NVIDIA GPU Display Driver for Windows and Linux allows a local attacker with low privileges to potentially achieve code execution, denial of service, privilege escalation, information disclosure, and data tampering. The flaw affects GeForce, RTX/Quadro/NVS, Tesla, and vGPU Manager products, and exploitation requires local access (CVSS AV:L) with no user interaction (UI:N) but low privileges (PR:L). No public exploit code has been identified at time of analysis, and there is no indication of active exploitation.
An out-of-bounds write in the kernel-mode layer of the NVIDIA GPU Display Driver for Windows and Linux allows a local, low-privileged authenticated user to corrupt kernel memory, with the potential to escalate privileges to kernel-level code execution, crash the system (denial of service), disclose information, or tamper with data. The flaw rates CVSS 3.1 7.8 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H), meaning exploitation requires pre-existing code execution as a standard user on a host with the driver loaded, but no user interaction and no non-default configuration; the vulnerable kernel-mode layer ships and loads by default once the driver is installed. There is no confirmed active exploitation (not in CISA KEV) and no public exploit identified at time of analysis, but the affected GeForce, RTX/Quadro/NVS, and Tesla product families are extremely widely deployed, making this a genuine high-priority patch-management item rather than an emergency.
Local attackers with low privileges can exploit an uninitialized pointer access flaw (CWE-824) in the firmware of NVIDIA Display Driver for Windows and Linux to achieve code execution, denial of service, privilege escalation, information disclosure, or data tampering. The vulnerability requires local access (AV:L) and authenticated low-privilege credentials (PR:L); no public exploit identified at time of analysis. Affected products include GeForce, RTX/Quadro/NVS, and Tesla product lines, including vGPU Manager deployments.
A use-after-free (CWE-416) in the NVIDIA GPU Display Driver for Windows and Linux lets a local unprivileged user free and then re-reference a driver-managed memory object, potentially yielding code execution in kernel context, privilege escalation, data tampering, denial of service, and information disclosure. The flaw is rated CVSS 3.1 7.8 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) and affects GeForce, RTX/Quadro/NVS, and Tesla product lines; the 'Guest Driver' tag indicates the vulnerable path may also be reachable from inside a virtualized guest, making guest-to-host or guest-context escalation a relevant scenario. Exploitation requires that the attacker already hold a local foothold with normal user privileges on a host with the affected driver installed and the GPU device interface exposed - there is no remote/network vector - and at time of analysis no public exploit code and no confirmed active exploitation have been identified, so this is a priority local privilege-escalation issue rather than an imminent, opportunistically exploited threat.
Local kernel stack memory disclosure in NVIDIA GPU Display Driver for Windows and Linux allows an attacker who already has low-privilege authenticated code execution on the host to read kernel stack contents such as return addresses and pointers. Affected installations include GeForce, RTX/Quadro/NVS, and Tesla driver configurations, and the independent assessment rates the flaw as AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N, meaning high confidentiality impact without confirmed integrity or availability impact. No public exploit is identified at time of analysis, and no CISA KEV active exploitation is reported.
An out-of-bounds read in the kernel-mode layer of the NVIDIA GPU Display Driver for Windows and Linux allows a local, low-privileged user to read beyond the bounds of kernel heap memory, producing information disclosure and denial of service, with potential for privilege escalation, code execution, and data tampering if the read primitive is successfully leveraged. The flaw affects GeForce, RTX/Quadro/NVS, and Tesla product families on systems with the vulnerable driver branch installed, and exploitation requires an attacker to already hold local access to the host (per the assessed CVSS:3.1/AV:L/AC:L/PR:L/UI:N vector, base 7.8), with no user interaction needed. No public exploit code has been identified at time of analysis, and the issue is not listed in CISA KEV, so exploitation is currently expected to require a motivated local attacker rather than being broadly weaponized.
An out-of-bounds read in the kernel-mode layer of the NVIDIA GPU Display Driver for Windows and Linux lets a local attacker who already holds a low-privileged execution context on the host read past the bounds of a kernel buffer, exposing kernel memory contents and creating a pathway toward further privilege escalation or data tampering. Exploitation is constrained to the local attack surface: the attacker must already be able to run code on the machine (AV:L, PR:L) and no user interaction or network path is involved, so despite the 7.8 headline score the practical risk is a local privilege/information-disclosure concern rather than a remotely reachable, mass-exploitable flaw. There is no confirmed active exploitation (not listed in CISA KEV) and no public exploit code identified at time of analysis; the affected hardware spans GeForce, RTX/Quadro/NVS, and Tesla products, with an NVIDIA vendor advisory available.
Local privilege escalation in the NVIDIA GPU Display Driver kernel-mode layer allows a user with standard privileges to trigger an out-of-bounds write in kernel memory, potentially achieving kernel-level code execution, denial of service, information disclosure and data tampering. The affected driver branches span NVIDIA GeForce, RTX/Quadro/NVS and Tesla families on both Windows and Linux, and the 'Guest Driver' tag indicates virtualized/vGPU and guest-VM driver deployments are also in scope, which matters for GPU-accelerated VDI and datacenter estates. Exploitation is constrained to the host: the CVSS 3.1 vector is AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8), so an attacker must already have a foothold and the ability to run code as a low-privileged local user; no public exploit code and no confirmed active exploitation were identified at time of analysis.
Local privilege escalation in NVIDIA GPU Display Driver for Windows and Linux allows a low-privileged local user to trigger an integer overflow in the kernel mode layer, resulting in an out-of-bounds write to GPU memory that can lead to code execution, denial of service, escalation of privileges, information disclosure, and data tampering. Affected platforms include GeForce, RTX/Quadro/NVS, and Tesla drivers, with the 'Guest Driver' tag indicating that VM guests with GPU access are also in scope for guest-to-host or in-guest escalation. No public exploit code or confirmed active exploitation (CISA KEV) has been identified at time of analysis, and the CVSS 3.1 score is 7.8 (AV:L/PR:L/UI:N).
Local privilege escalation in NVIDIA GPU Display Driver for Windows and Linux allows a low-privileged user to trigger an incorrect conversion between numeric types in the kernel mode layer, potentially leading to code execution, denial of service, escalation of privileges, information disclosure, and data tampering. The vulnerability requires local access and low-level privileges (CVSS vector PR:L) with no user interaction, and it is not remotely exploitable. No public exploit code has been identified at time of analysis, but the risk is materially higher on shared or multi-tenant GPU hosts such as VDI, vGPU, or Tesla datacenter deployments where untrusted users can reach the driver.
Local privilege escalation in the NVIDIA GPU Display Driver for Windows and Linux lets a low-privileged local user trigger an out-of-bounds array access in the kernel mode layer, potentially yielding code execution in kernel context, denial of service, escalation of privileges, information disclosure and data tampering. The assessed CVSS 3.1 vector is AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8), meaning the attack is strictly local and requires only low-privileged authenticated code on the host, with no user interaction and low complexity - so any unprivileged process able to reach the affected kernel-mode interface on a system running GeForce, RTX/Quadro/NVS or Tesla drivers (or a vGPU host reachable from a guest) can attempt it. There is confirmed actively exploited (CISA KEV) status: no - the CVE is not in KEV and no public exploit code has been identified at time of analysis, so early patching is key while exploitation activity has not yet been observed.
Local code execution, privilege escalation, denial of service, information disclosure, and data tampering are possible on Linux systems running vulnerable NVIDIA GPU Display Drivers (GeForce, RTX/Quadro/NVS, and Tesla branches) because of a type-confusion flaw (CWE-843) in the outdated embedded cryptographic library used by the NGX updater. The assessed profile is CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8): an attacker who already holds a low-privileged local account can reach the vulnerable updater code path, with no user interaction required and low complexity, but the flaw is not remotely reachable, so shared and multi-user hosts (VDI, HPC/Tesla compute clusters, and workstations with untrusted local users) are the realistic risk surface. No public exploit code has been identified at time of analysis, and the exact affected and fixed driver versions must be confirmed from NVIDIA's advisory.
Integer overflow in the kernel-mode layer of NVIDIA vGPU Virtual GPU Manager can be triggered by a low-privileged user inside a guest VM, causing memory corruption with high confidentiality, integrity, and availability impact. The vulnerability affects vGPU deployments on Windows and Linux hosts across GeForce, RTX/Quadro/NVS, and Tesla driver branches, and successful exploitation could enable code execution, denial of service, privilege escalation, information disclosure, or data tampering. No public exploit code or active exploitation has been identified in the provided intelligence at time of analysis.
Local privilege escalation in the NVIDIA GPU Display Driver for Linux allows an authenticated, low-privileged user to drive an out-of-bounds write in the driver's kernel-mode layer by supplying mismatched memory buffers during event buffer setup, with a successful exploit potentially leading to kernel code execution, denial of service, privilege escalation, information disclosure, and data tampering (CVSS 3.1 base 7.8, AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H). The affected code is the Linux kernel-mode component shipped with GeForce, RTX/Quadro/NVS and Tesla driver packages, and the same path can be reached by untrusted code inside a guest VM using the NVIDIA vGPU guest driver on a host running the Virtual GPU Manager. No public exploit code was identified and there is no confirmed active exploitation (CISA KEV) at time of analysis; the bug is local-only, requiring the attacker to already run code with at least low privileges on the host or in a guest, and it is not remotely reachable.
Local privilege escalation in the NVIDIA GPU Display Driver kernel-mode layer lets an attacker who already holds low-privileged code execution on a Windows or Linux host - including code running inside a vGPU guest VM - drive the driver into an error path that mishandles reference counts, producing a use-after-free. The corrupted object is kernel memory, so successful exploitation can yield code execution, escalation of privileges, denial of service, information disclosure, and data tampering (CVSS 7.8, AV:L/AC:L/PR:L/UI:N); it is not remotely reachable and requires a vulnerable driver version to be installed. The GeForce, RTX/Quadro/NVS, and Tesla driver families are in scope, and no public exploit code has been identified at the time of analysis.
Local privilege escalation in the NVIDIA GPU Display Driver for Linux lets an already-authenticated, low-privileged user trigger a format string defect (CWE-134) and potentially achieve code execution, privilege escalation, data tampering, denial of service, and information disclosure; NVIDIA rates it 7.8 (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H). The attack surface is strictly local - there is no remote vector - so real risk concentrates on multi-user and multi-tenant GPU systems such as HPC clusters, ML/AI training nodes, shared engineering workstations, and cloud GPU instances, where a low-privileged tenant could reach the driver while other users' data and the host remain in scope. No public exploit code has been identified at time of analysis and the CVE is not listed in CISA KEV; the impact rating of total confidentiality, integrity, and availability loss is consistent with memory corruption inside a privileged driver context.
Local privilege escalation and kernel memory-safety compromise on Linux systems running NVIDIA GPU Display Drivers allow an unprivileged local user to cause improper release of kernel-mode memory resources, leaving a mapping accessible after the underlying memory has been reused. The flaw affects GeForce, RTX/Quadro/NVS, and Tesla driver lines and carries a CVSS 3.1 base score of 7.8 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H), meaning an attacker who already holds a local login can potentially achieve code execution, denial of service, information disclosure, privilege escalation, and data tampering without user interaction. No public exploit has been identified at time of analysis and the issue is not confirmed as actively exploited, but the local-only vector makes it most dangerous on shared or multi-tenant GPU hosts where one unprivileged tenant could compromise the kernel and other workloads.
Local privilege escalation and memory-protection bypass in NVIDIA GPU Display Driver for Linux allows an unprivileged local user to execute code at higher privilege, tamper with data, disclose information, or cause denial of service. The flaw is in the kernel-mode layer and affects GeForce, RTX/Quadro/NVS, Tesla, and vGPU Guest Driver/Virtual GPU Manager deployments, with a CVSS 3.1 score of 7.8 (AV:L/AC:L/PR:L/UI:N). No public exploit identified at time of analysis, and no CISA KEV active-exploitation signal is present; the assessed attack path requires prior local access on the host or inside a vGPU guest.
Container escape in CRI-O - the OCI runtime that underpins Red Hat OpenShift Container Platform 4 - is achievable when a workload-level actor can create or influence pod metadata that overwrites CRI-O's own reserved sandbox bookkeeping; the poisoned state is then reloaded as trusted after a CRI-O daemon restart, and a subsequent container recreate inside that sandbox can expose a host-side runtime-management resource to the container. Rated CVSS 3.1 7.8 (AV:L/AC:H/PR:L/UI:N/S:C/C:H/I:H/A:H), the flaw is not remotely reachable without cluster workload access, and the mandatory restart-then-recreate sequencing is the principal limiter on realistic exploitation. No public exploit code has been identified at time of analysis and the issue is not confirmed actively exploited (no CISA KEV listing); no EPSS score was supplied in the available intelligence.
Memory corruption in U-Boot's DHCPv6 netboot path allows an attacker on the same local network to crash or corrupt the bootloader by sending crafted DHCPv6 ADVERTISE or REPLY packets with oversized DUID (SERVERID/CLIENTID) option lengths. All U-Boot builds prior to 2026.10-rc5 that perform DHCPv6-based network boot are affected; the assessed impact is availability-focused (CVSS:3.1 AV:A/AC:L/PR:N/UI:N/C:N/I:L/A:H), with confidentiality unaffected and integrity impact limited. Exploitation requires adjacency to the booting device and only works during the brief netboot window, so devices that boot from local storage are not exposed; no public exploit and no confirmed active exploitation were identified at time of analysis.
A use-after-free in U-Boot's lwIP-based wget HTTP client lets an attacker who controls (or MITMs) the server a device downloads from crash the bootloader during a netboot/provisioning fetch, interrupting boot. The bug sits in the httpc_recv_cb() receive callback: when HTTP data storage fails, it frees the connection's protocol control block but returns ERR_BUF instead of ERR_ABRT, so the TCP input path continues on released memory. All U-Boot versions before 2026.10-rc5 are affected; the CVSS 4.0 vector scores 8.8 (AV:N/AC:L/PR:N/UI:N, VA:H, VI:L) and our independent CVSS 3.1 assessment rates it 7.5 as an availability-dominant issue - no public exploit identified at time of analysis, and impact is limited to a bootloader crash rather than code execution or data exposure.
Memory corruption in U-Boot’s NFS READLINK handling can crash the bootloader on devices booting over NFS. A malicious or man-in-the-middle NFS server can return a crafted symlink length that triggers a signed-to-unsigned conversion error (CWE-195) and out-of-bounds access in nfs_readlink_reply(), affecting U-Boot versions before 2026.10-rc5; the vendor CVSS 4.0 base score is 8.8. Exploitation requires the device to be actively performing an NFS netboot and reaching the attacker-controlled server, so idle or non-NFS-boot devices are unaffected, and no public exploit code or CISA KEV active exploitation is identified at time of analysis.
Memory corruption in the U-Boot bootloader's NFS client lets a malicious or man-in-the-middle NFS server overflow the packet buffer in nfs_read_reply(), crashing the bootloader or corrupting memory before control is handed to the kernel. Affected are U-Boot builds prior to 2026.10-rc5 that perform NFS netboot (CPE cpe:2.3:a:u-boot:u-boot:*); the flaw is reachable by an unauthenticated network attacker (CVSS vector PR:N), though per our assessment the practical outcome is a boot-time denial of service with limited integrity impact rather than confirmed code execution, since the device must be actively NFS-booting and the attacker must control or intercept the NFS server on the boot network. A vendor patch exists (commit 0bbf0985, fixed in 2026.10-rc5) and no public exploit code was identified at time of analysis.
Memory corruption in U-Boot's IP fragment reassembly code (net/net.c) allows remote, unauthenticated attackers on the boot network to crash the bootloader and corrupt adjacent memory while a device is performing a netboot. Only builds compiled with CONFIG_IP_DEFRAG are affected - an opt-in, frequently disabled compile option - so deployments without it are not vulnerable, and exploitation requires delivering a crafted IP fragment with a non-zero offset and the More-Fragments flag onto the boot path during the narrow netboot window. The CVSS 4.0 base score is 8.8 (CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:L/VA:H), an availability-dominant flaw with only limited integrity impact and no confidentiality loss, and no public exploit has been identified at time of analysis (no CISA KEV listing).
Authentication bypass in the optional sshd-ldap component of Apache MINA SSHD (versions 1.2.0 through 2.19.0 and 3.0.0-M1 through 3.0.0-M5) lets unauthenticated remote attackers complete SSH authentication by submitting the literal username "*" with the password "*", because LDAP filter metacharacters from user-supplied credentials are not escaped before being embedded in the search filter. Rated CVSS 9.1 (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N), the flaw is trivially reachable but only on servers that both embed the MINA SSHD library and have deployed and configured sshd-ldap for password or publickey authentication; stock MINA SSHD deployments without that optional component are entirely unaffected, and no public exploit code was identified at time of analysis. Apache ships fixes in 2.20.0 and 3.0.0-M6, which apply proper RFC 4515 escaping.
Authentication bypass in the optional sshd-ldap component of Apache MINA SSHD allows unauthenticated remote attackers (CVSS:3.1/AV:N/AC:L/PR:N/UI:N) to pass password authentication checks when an SSH server is configured to use LdapPasswordAuthenticator, yielding high confidentiality and integrity impact (rated 9.1 critical). Affected builds are 1.2.0 through 2.19.0 and 3.0.0-M1 through 3.0.0-M5; servers that rely on sshd-core's built-in password mechanisms, or that do not use sshd-ldap at all, are not affected. No public exploit code has been identified and the issue is not confirmed actively exploited (no CISA KEV listing), but real-world exposure is sharply limited by the non-default optional component and the explicit LDAP authenticator configuration required - the raw 9.1 score is driven by impact rather than by a broad exploitable population.
Memory exhaustion in the Apache MINA SSHD SFTP client (DefaultSftpClient, sshd-sftp component) lets an attacker-controlled SFTP server drive a client application to run out of heap and crash, because the client stores unsolicited SFTP replies that it never matches to an outstanding request and therefore never releases. The flaw affects client-side deployments of Apache MINA SSHD versions 0.9.0 through 2.19.0 and 3.0.0-M1 through 3.0.0-M5; server-only deployments are not exposed, and the victim must initiate an outbound SFTP session to a malicious or compromised server. It is rated CVSS 3.1 7.5 (AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H) and is a pure denial-of-service condition of the client process, with no authentication bypass, code execution, or data disclosure; no public exploit code was identified at time of analysis and the issue is not listed as confirmed actively exploited (CISA KEV).
A partial authentication bypass in Apache MINA SSHD lets a remote attacker satisfy a repeated-public-key multi-authentication policy by presenting the same public key twice, so a server configured to demand two distinct keys can be entered with only one of them. Affected are versions before 2.20.0 and 3.0.0-M1 through 3.0.0-M5 of the sshd-core component; the vendor assigns CVSS 3.1 8.1 (High, CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:N), where PR:L reflects that the attacker must already legitimately possess at least one accepted key pair. Exploitation is confined to the non-default hardened configuration that requires repeated/multiple public keys (the equivalent of OpenSSH's AuthenticationMethods "publickey,publickey"); single-key or password-based deployments are not affected, and no public exploit code or confirmed active exploitation was identified at time of analysis.
An authentication bypass in Apache MINA SSHD's sshd-core library (versions 2.0.0 through 2.19.0 and 3.0.0-M1 through 3.0.0-M5) allows remote, unauthenticated attackers to skip signature verification during public-key or hostbased SSH authentication, potentially granting an attacker a valid session. The CVSS 3.1 vector AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N (9.1 critical) reflects high confidentiality and integrity impact, but exploitation is gated by a non-default precondition: the server operator must have written explicit custom code enabling the library's 'asynchronous authentication' mechanism and applied it to public-key or hostbased authentication. Apache rates this implementation choice as presumed rare, and servers that do not opt into asynchronous authentication - or that use it only with password or keyboard-interactive auth - are not exposed. No public exploit code has been identified at time of analysis and the vulnerability has not been confirmed as actively exploited (no CISA KEV listing).
Signed integer overflow in the s2 dictionary API of klauspost/compress before 1.18.7 lets an attacker crash a consuming Go process: a dictionary carrying a uvarint-encoded repeat value larger than MaxInt64 slips past repeat-index validation, and a subsequent call to Dict.Encode() converts the overflowed negative repeat value into an out-of-bounds unsafe.Pointer access that aborts the process with SIGSEGV. The issue is availability-only (C:N/I:N, A:H) on a CVSS:3.1/AV:N/AC:L/PR:N/UI:N vector, but real exposure hinges on the application - it only matters where untrusted, attacker-influenced bytes reach s2.NewDict() and the resulting dictionary is then encoded, not in ordinary compression or decompression of arbitrary data. No public exploit code and no CISA KEV entry were identified at time of analysis; the vendor-released patch is v1.18.7.
Heap-based out-of-bounds write in the RPM Package Manager (rpm) affects Red Hat Enterprise Linux 6 through 10 and Red Hat Hardened Images, allowing a local, unauthenticated attacker to corrupt memory by crafting an RPM package that is then parsed with rpm2cpio, rpm2archive, or rpm -qlvp. The flaw arises from an integer overflow in iterReadArchiveNext() when a symlink entry declares RPMTAG_LONGFILESIZES as 0xFFFFFFFFFFFFFFFF, shrinking the buffer allocation to one byte and enabling attacker-controlled data from the cpio filesize field to overwrite heap memory. Exploitation requires user interaction to inspect the malicious package and no public exploit code has been identified at time of analysis; the CVSS 3.1 base score is 7.1.
Local privilege escalation in Ghostscript for Windows (all versions prior to 10.08.0) allows any authenticated local user to plant a malicious PostScript resource file in the predictable C:\gs\ path, which Ghostscript then auto-loads with the privileges of whichever principal next invokes it. The hijack requires no special Ghostscript configuration because it abuses Windows default ACLs that let authenticated users create directories at the root of C:\, and it succeeds only while C:\gs\ is unpopulated - the default state on Windows installations; execution is passive and fires when a user or service subsequently runs Ghostscript, with the payload's privilege gain bounded by that invoking principal (assessed as scope-changed impact). No public exploit code has been identified at time of analysis, and the issue, reported by CERT-PL, is fixed in 10.08.0.
Malformed SCTP traffic can crash the Wireshark/tshark dissector in versions 4.6.0 through 4.6.8 and 4.4.0 through 4.4.18, turning a heap-based buffer overflow (CWE-122) in the SCTP dissector into a denial of service against the analyst's own capture session rather than against any remote service. Exploitation requires the tool to process an attacker-supplied packet - either live-captured on a monitored interface carrying crafted SCTP traffic or opened from a malicious capture file - and the packet must be specifically malformed, which is why the high attack-complexity rating applies. Access requirements are unauthenticated per the CVSS vector (PR:N), and the realistic outcome is a crash of the Wireshark/tshark process, not code execution or host compromise; no public exploit code and no confirmed active exploitation were identified at time of analysis. Note that the originally supplied CVSS vector claimed high confidentiality and integrity impact, which the independent assessment corrected to availability-only (C:N/I:N/A:H).
Wireshark 4.6.0 through 4.6.8 and 4.4.0 through 4.4.18 can be crashed by processing attacker-controlled SPDY protocol data, an availability-only denial-of-service condition that our independent assessment scores as CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:L. The signals conflict and, on balance, point to a LOWER real-world priority than the headline 8.8 suggests: exploitation requires the victim to open a crafted capture file or live-capture crafted SPDY traffic, publicly available exploit code exists, EPSS is only 0.45% (37th percentile), and there is no confirmed active exploitation (CISA KEV). Unauthenticated attackers can trigger the crash, but user interaction is mandatory and the impact is limited to terminating the Wireshark process, not code execution or data compromise.
Path traversal in Wireshark's configuration-profile import handler (CWE-22) allows a maliciously crafted profile to write files outside the intended profile directory, crashing the application and potentially achieving code execution in the user's context. Affected releases are Wireshark 4.4.0 through 4.4.18 and 4.6.0 through 4.6.8; the flaw is rated CVSS 7.8 (AV:L/AC:L/PR:N/UI:R) and requires the victim to manually import the attacker-supplied profile, so exploitation depends on social engineering rather than a network-facing or automatic trigger. Publicly available exploit code exists and the issue is tracked as wnpa-sec-2026-106, but it is not confirmed as actively exploited (CISA KEV) and EPSS is low at 0.19% (7th percentile), making this a moderate-priority issue rather than an emergency; patched releases 4.4.19 and 4.6.9 are available from the vendor.
Use-after-free in libpng 1.6.0 through 1.6.58 can crash applications that use the sequential reader and call png_read_end after png_read_info without first reading the image rows. A crafted PNG containing a zTXt, iTXt or iCCP chunk that makes libpng abandon zlib decompression early leaves a stale input pointer in the zlib stream; the zTXt/iTXt path then reads through a freed heap buffer (a larger tEXt or pCAL chunk before IDAT must first reallocate the read buffer), while the iCCP path is a stack use-after-return. Impact is confined to denial of service - the dangling pointer is only read into a discarded scratch buffer, so there is no data disclosure or corruption - and the required API call sequence is rare in practice; no public exploit has been identified at time of analysis, and the vendor rates it 5.9 Medium (CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H), consistent with the independent assessment.
Signature verification bypass in Authlib 1.7.2 and earlier allows unauthenticated remote attackers to have crafted JWS JSON Serialization payloads accepted as valid, because JsonWebSignature.deserialize_json() returns the payload without checking for a signature or requiring a cryptographic key. The flaw affects applications that use that method to process incoming tokens and then trust the returned payload for authentication, authorization, or integrity decisions; applications using only compact JWS or independently verifying signatures are not exposed. No public exploit code has been identified at time of analysis, and EPSS is low at 0.12% (2nd percentile), but the issue is remotely exploitable with no authentication or user interaction and is assessed as high severity.
Authentication bypass in the LiteLLM proxy lets any party who already holds a valid JWT issued by the configured external identity provider impersonate an existing user - including accounts with proxy_admin privileges - by presenting a token whose unverified email claim matches the victim's address. The JWT authentication flow performs an email-based fallback lookup without validating the email_verified claim, so the attacker inherits the victim's role and can permanently overwrite the victim's stored identity binding, retaining persistent access to administrative endpoints that expose API keys and user management. The attacker must possess a legitimately signed token from the same IdP (CVSS PR:L, AC:H), deployments that do not use JWT/IdP authentication are unaffected, and publicly available exploit code exists; no confirmed in-the-wild exploitation has been reported.
Type confusion in XS::Parse::Infix versions 0.40 through 0.49 - the Perl XS module distributed as part of XS-Parse-Keyword - lets an attacker who can influence a list handed to a list-associative infix operator (for example the zip operator from Syntax::Operator::Zip) make the interpreter dereference an address of the attacker's choosing. Supplying a bare number where an array reference is expected reliably segfaults the process (denial of service), while a crafted string carrying 0x0b at byte 12 slips past the faulty reference guard and is treated as an AV head, so bytes 16-23 are used as AvARRAY and its entries are pushed onto the Perl stack as live SVs, yielding a limited out-of-bounds read / information-disclosure primitive. Publicly available exploit code exists (a minimal zip-based proof-of-concept is described by the reporter), but there is no confirmed active exploitation (no CISA KEV listing) and EPSS is low at 0.18% (7th percentile); this is a library flaw with no standalone attack surface, so real-world risk depends entirely on whether a consuming application passes untrusted, attacker-influenced data into such an operator.
Stored cross-site scripting in the SUSE Rancher UI allows unauthenticated remote attackers to inject malicious script through updates to public UI settings, which then executes in the browser of any user-including administrators-who later views the affected interface. Affected versions are SUSE Rancher 2.15 before 2.15.2, 2.14 before 2.14.6, 2.13 before 2.13.10, 2.12 before 2.12.14, and 2.11 before 2.11.18. Vendor CVSS 3.1 rates this 9.6, while our independent assessment lowers the availability impact to low; no public exploit identified at time of analysis, and exploitation requires network reachability to the Rancher management endpoint plus a victim loading the UI.
Remote attackers holding a previously issued Rancher session token can retain access even after the legitimate user logs out in SUSE Rancher 2.15 before 2.15.2, because the server fails to clear credentials on logout (CWE-613). This session persistence flaw carries a CVSS 3.1 base score of 8.1 (High) with network reach, low complexity, no attacker privileges (PR:N), and required user interaction (UI:R) for the logout. No public exploit identified at time of analysis; exploitation is limited to scenarios where the attacker already possesses a valid token, such as shared-device or stolen-session situations.
Privilege escalation in SUSE Rancher Fleet (0.14.x before 0.14.11, 0.15.x before 0.15.7, and 0.16.x before 0.16.2) lets a low-privileged tenant - an authenticated user able to push a Fleet Bundle or GitRepo - have resources written to a shared downstream cluster under the Fleet agent's own cluster-admin credentials rather than the ServiceAccount pinned to the deployment, allowing that tenant to overwrite configuration owned by other tenants. Exploitation is gated by topology: it only matters in multi-tenant Fleet deployments where privileged and untrusted teams share the same downstream cluster, since single-tenant or per-tenant-isolated clusters have no crossed trust boundary. Rated 8.8 by the vendor (the independent assessment scores it higher, 9.9, because the trust boundary crossing changes scope); no confirmed active exploitation (CISA KEV) and no public exploit identified at time of analysis.
Path traversal in Flatpak's app deployment handling of the exported files/etc directory allows a malicious Flatpak application to empty or replace host files such as /etc/passwd, /etc/group, /etc/machine-id, or /etc/resolv.conf with a symlink when that app is installed or upgraded; on system-wide installations the write is carried out as root. Exploitation is gated on user interaction: the victim must choose to install or upgrade the attacker-supplied app from an untrusted source (CVSS 3.1 base 7.1, AV:N/AC:L/PR:N/UI:R with high availability and low integrity impact), making this a genuine but moderate-priority integrity/availability issue rather than a critical remote code execution flaw. No public exploit code and no confirmed active exploitation (CISA KEV) were identified at time of analysis.
Path traversal in Flatpak's handling of the export/bin directory during application deployment allows a malicious Flatpak app to delete attacker-chosen files outside the deployment directory when that app is installed or upgraded; on system-wide installations the deletions are carried out with root privileges, so a crafted package can remove arbitrary files anywhere the deployer can reach. The flaw affects Flatpak releases before 1.18.4 and is triggered by a user or administrator installing or upgrading an attacker-supplied app, which supplies the required user interaction rather than a network-based trigger; the CVSS vector carries PR:N, so no prior authentication is needed beyond that install action. No public exploit code has been identified at time of analysis and there is no evidence of confirmed active exploitation, though the integrity and availability impact is high (arbitrary file deletion) and the flaw crosses a privilege boundary on system-wide installs.
Cross-tenant authorization bypass in SUSE Rancher Fleet allows a party who can perform agent-initiated cluster registration into a Fleet workspace namespace shared with other tenants to inject attacker-controlled cluster labels - including reserved management.cattle.io/ labels such as the cluster display name - that are then persisted onto the upstream Cluster object. Because Fleet resolves GitRepo and Bundle targeting from those labels, the attacker's own cluster can be made to satisfy targeting rules administrators intended for a different tenant's cluster, causing content such as Git repos and bundles to be deployed to the wrong cluster. The flaw affects Fleet 0.16 before 0.16.1, 0.15 before 0.15.6, 0.14 before 0.14.10, 0.13 before 0.13.15, 0.12 before 0.12.19 and older releases; exploitation is conditional on a multi-tenant deployment with a shared workspace namespace and label-based targeting, requires a low-privileged authenticated position (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:N, 7.1), and no public exploit code has been identified at time of analysis.
Path traversal in the btrfs storage driver of Canonical LXD 4.0.2 and later allows an authenticated client that holds project-scoped permission to create instances to delete arbitrary files on the host as root by supplying a crafted subvolume path containing ../ sequences, and on hosts whose root filesystem is itself btrfs the same primitive escalates to placing attacker-controlled content at arbitrary host paths for full host compromise. The issue is rated CVSS v3.1 9.6 (AV:N/AC:L/PR:L/UI:N/S:C/C:N/I:H/A:H), reflecting a scope-changing breakout from a restricted tenant role into the host root context rather than a CVSS-inflated score, since a legitimate but limited project credential is the only prerequisite. No CISA KEV listing and no public exploit code were identified at time of analysis; vendor fixes are available in LXD 4.0.14, 5.0.10, 5.21.8 and 6.10.
Path traversal in Canonical LXD's Btrfs storage driver allows an authenticated user with instance-creation privileges to escape the intended backup import target and delete or replace arbitrary files and directories on the host as root, by supplying a crafted subvolumes[].path value in backup/optimized_header.yaml during a btrfs 'optimized' backup import. The assessed vector (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H, 9.9) reflects high confidentiality, integrity and availability impact with a scope change because the traversed write lands outside the LXD daemon's own container boundary; exploitation requires valid low-privilege API credentials and a Btrfs-backed host, not anonymous access. No public exploit code has been identified at time of analysis, and the issue is not confirmed actively exploited (CISA KEV), but the low attack complexity and root-level file-system effect make it a genuine high-priority fix rather than a CVSS-inflated score.
OS command injection in SUSE NeuVector 5.4 (before 5.4.11), 5.5 (before 5.5.4), and 5.6 (before 5.6.2) allows an attacker who already satisfies one of two prerequisites - an authenticated NeuVector user holding the namespaced Runtime Policies (write) permission, or possession of NeuVector's internal gRPC certificate key pair - to inject operating-system commands that execute inside the privileged enforcer container, resulting in complete compromise of the underlying worker node. The flaw is rated 8.8 (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H), meaning remote reachability, low attack complexity, authenticated low privileges, and no user interaction, with total confidentiality, integrity, and availability impact once the prerequisite is met. Ordinary read-only users and unauthenticated network clients cannot trigger it, but any deployment that grants Runtime Policies write access broadly or leaks the internal gRPC key pair is at high risk; no public exploit code has been identified at time of analysis and the issue is not listed in CISA KEV, though weaponization would be straightforward given the trivially reachable command sink.
An out-of-bounds read in libXtst's RECORD reply parser (CWE-126) allows a malicious or compromised X server to crash attached X clients that link the library and use the XTEST/RECORD extension, affecting all versions before 1.2.6. The assessed vector is CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:L/I:N/A:L, meaning unauthenticated attackers in the network position of the rogue server can trigger a low-integrity information leak and a denial of service, but only if the victim voluntarily connects to that server (UI:R). This is a reverse-trust scenario: it is not exploitable against clients that connect only to trusted local or remote X servers, and the X server itself is unaffected. There is no public exploit identified at time of analysis and the issue is not in CISA KEV; vendor patch data indicates a fix is available, with the upstream correction landing in the libXtst 1.2.6 line.
A race condition in the Linux kernel cgroup task iterator (css_task_iter_next) can hand back a task_struct whose reference count has already dropped to zero, resulting in a use-after-free that a local attacker can turn into kernel memory corruption. Only systems running affected kernels built with cgroups are exposed, and the attacker must be able to read a cgroup.procs file while concurrently spawning and tearing down multi-threaded process groups - a low-privileged but purely local position (CVSS:3.1/AV:L/AC:H/PR:L). Exploitation is gated by an extremely narrow timing window tied to RCU grace periods and task teardown ordering, so the flaw is high-severity on paper but difficult to trigger reliably; EPSS is only 0.18% (7th percentile) and no public exploit code was identified at time of analysis.
Path traversal via symlink prefix reuse in gitoxide's gix-fs crate (before 0.23.0; GHSA lists <= 0.21.0) allows a crafted repository tree to write attacker-controlled symlinks outside the worktree during forced checkout with overwrite_existing enabled, enabling code execution by overwriting known paths such as .git/hooks/post-checkout. Exploitation is local, requires user interaction to initiate checkout of an untrusted tree, and has high attack complexity tied to winning the prefix-reuse condition; no public exploit was identified at time of analysis, and no KEV or EPSS data were provided. Per the assessed CVSS:3.1 vector AV:L/AC:H/PR:N/UI:R/S:U/C:H/I:H/A:H, unauthenticated but local access with user interaction is required.
Reflected cross-site scripting in the RabbitMQ management UI lets an attacker execute arbitrary JavaScript in the management interface origin by injecting content into a bearer token that is embedded unescaped in the OAuth bootstrap script generated by set_token_auth/2. RabbitMQ server 4.2.0 through 4.2.7 and 4.3.0/4.3.1 are affected only when the management plugin is configured with management.oauth_enabled=true, which is what exposes the OAuth bootstrap endpoint before authentication; the attacker must also induce an administrator to load the affected page (UI:R), and the cookie-based variant additionally requires the attacker to first plant an access_token cookie on the management host. The vendor-supplied CVSS 4.0 base score of 7.5 is higher than our independent CVSS 3.1 assessment (AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:L/A:N), which better reflects the non-default configuration gate and the extra cookie-planting prerequisite; no public exploit code has been identified at time of analysis.
A tenant isolation bypass in BerriAI LiteLLM versions before 1.101.0-rc.1 lets an authenticated holder of a valid virtual key read other tenants' entries from the semantic cache and, in the worst case, have attacker-supplied tool calls executed under a victim's credentials. The flaw stems from an inconsistency in how the cache layer derives the tenant scope versus the metadata variable name, so semantically similar prompts sent to routes such as /v1/responses and /bedrock/* can collide with and return another tenant's cached response - potentially exposing PII, financial data, or source code. Exploitation requires a multi-tenant deployment with semantic caching enabled and a sufficiently similar prompt to hit the victim's cache entry, and it is rated CVSS 8.7 with a vendor patch already released (commit 16db51e2, fixed in v1.101.0-rc.1); no public exploit code has been identified at time of analysis and the CVE is not in CISA KEV.
A use-after-free race condition in QEMU's 9pfs (virtfs) passthrough subsystem allows a malicious guest user to escape the shared directory boundary and read or write arbitrary files on the host, potentially achieving code execution as the QEMU process user (VM escape). Only hosts that have enabled 9pfs/virtfs host-directory sharing are exposed, since this is not a default configuration; the attacker must be a local guest user able to issue 9p protocol requests and must win a race between QEMU's main thread and a 9pfs worker thread with concurrent Tlcreate and Twalk requests. No public exploit code or active exploitation telemetry was identified at time of analysis, but the flaw is rated high severity (CVSS 8.8, scope-changed, C:H/I:H/A:H) and affects QEMU packages shipped in Red Hat Enterprise Linux 6 through 10, Red Hat OpenShift Container Platform 4, and related Red Hat products.
Command injection in the kitty terminal emulator (versions 0.47.3 through 0.48.x) lets any program that can get output rendered on screen - a file shown with cat, a log line, command output, an SSH banner, or a web response - make kitty write attacker-composed printable ASCII back into the pseudoterminal, where the user's shell treats it as typed input and executes it with the user's privileges. The root cause is that the OSC 21 colour-control handler echoed unrecognised field names verbatim into its reply, and the 0.47.3 restriction (introduced as the fix for CVE-2026-54057) to printable ASCII excluding ';' and '=' still left enough characters to compose a shell command; the newline needed to submit that command is obtainable through the ssh kitten path. Exploitation is local (AV:L) and requires the victim to display attacker-controlled bytes (UI:P), no privileges are required, and no public exploit code has been identified in the available intelligence at time of analysis. The 9.3 CVSS 4.0 base score overstates real-world reach, since this is not a remote, unauthenticated, one-click compromise; the vendor-released patch is kitty 0.49.0.
The Linux kernel's nvme-rdma host driver mishandles command teardown when a queue_rq submission returns -EIO, unmapping SQE DMA and cleaning up the block-layer request after the host-path-error helper has already completed that request. The result is a double cleanup of the nvme-rdma command and blk-mq request plus a DMA unmap against an already-completed request, which can corrupt request/command state, trigger a kernel crash or denial of service, and - per the assessed vector - cause limited integrity impact. Only hosts running the nvme-rdma driver while connected to an NVMe-oF target over an RDMA fabric (RoCE, iWARP or InfiniBand) are exposed, and the faulting branch only executes on a transient RDMA command-submission failure that is not directly attacker-controllable; the local, high-complexity, privileged nature of the trigger makes real-world risk substantially lower than the base score suggests, and no public exploit identified at time of analysis.
Improper CPU-ID validation in the Linux kernel's BPF subsystem (the BPF_F_CPU map-operation flag) allows a local user with BPF privileges - CAP_BPF, or any environment where unprivileged BPF is enabled - to drive an out-of-bounds per-CPU map access on hosts whose possible-CPU mask is sparse/non-contiguous, producing a kernel paging fault and denial of service (assessed availability impact A:H, with only limited integrity impact and no confidentiality impact). Only kernels exposing the relatively new BPF_F_CPU flag (introduced around Linux 7.0) are in scope, and exploitation additionally requires a machine with a CPU ID gap such as 0,2-3 - typical of certain arm64/embedded boards or virtualized guests with CPU device-tree holes - while mainstream x86 systems with contiguous 0..N-1 CPUs are not affected because num_possible_cpus() equals nr_cpu_ids there. There is no remote vector, exploitation is local and requires BPF privileges (CVSS vector PR:L, AC:H), and no public exploit code was identified at time of analysis; no CISA KEV entry was provided, so active exploitation is not confirmed, and this is best characterized as a high-CVSS-but-low-real-risk issue given the sparse-mask and CAP_BPF prerequisites.
Incorrect interpretation of the NPU_SET_IFM/OFM_PRECISION field by the Linux kernel's accel/ethosu driver on Arm Ethos-U65 hardware lets a local user with access to the accelerator character device corrupt command-stream processing. Because bits 15:14 are read as an activation storage mode instead of the U65 rounding mode, natural rounding (value 2) streams are rejected with -EINVAL, and truncating rounding (value 1) streams are misrouted to the two-tile address path where a feature-map size is computed from tile bases the stream never initialized, degrading integrity of NPU operations and potentially crashing the affected subsystem. Authentication is required (CVSS vector PR:L, local vector AV:L); no public exploit code has been identified at time of analysis, and exploitation is limited to systems running the driver bound to real Ethos-U65 silicon, such as NXP i.MX93 SoCs. Vendor-released patches are available in the referenced kernel stable commits, with EUVD listing Linux 7.2.7 and 7.3-rc2 as fixed.
Linux kernel SCTP timer handling is vulnerable to a time-of-check-to-time-of-use race that can unbalance an association's reference count and cause a use-after-free during association teardown, allowing an unauthenticated remote peer to crash or destabilize an affected system. Any kernel built with the SCTP protocol enabled (CONFIG_IP_SCTP) and exposing an SCTP endpoint to untrusted peers is in scope; the assessed vector is AV:N/AC:H/PR:N/UI:N with no confidentiality impact, low integrity impact, and high availability impact, because winning the narrow timing race is the attacker's main obstacle. There is no public exploit identified at time of analysis and no CISA KEV listing; the upstream and stable-branch fixes are already published, and no EPSS score was supplied in the available intelligence.
Use-after-free in the Linux kernel's VXLAN multicast database (MDB) handling lets a local, low-privileged attacker corrupt kernel memory by crafting netlink MDB requests, though no public exploit code was identified at time of analysis and the issue is not confirmed as actively exploited (CISA KEV). The root cause is that vxlan_mdb_is_valid_source() accepts the all-zeros address as both MDBE_ATTR_SOURCE and a source-list member, causing the derived (S,G) key to be byte-identical to the (*,G) key, so vxlan_mdb_remote_src_del() frees an entry and then deletes it a second time. The independently assessed impact is limited to integrity and availability (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:L/A:H) rather than the higher confidentiality/integrity scores in the vendor-published vector, because the flaw is a local write primitive and denial-of-service/crash risk rather than a disclosure channel; exploitation requires a user namespace where an unprivileged user can obtain CAP_NET_ADMIN in a new network namespace, and is blocked where unprivileged user namespaces are disabled.
A page-level use-after-free in the Linux kernel ALSA PCM subsystem lets a local user with access to a /dev/snd PCM character device turn a race between a concurrent mmap() and a buffer reallocation into a stale, writable mapping of freed DMA buffer pages, which can be reclaimed by the allocator and reused for local privilege escalation. The attacker must issue SNDRV_PCM_IOCTL_HW_PARAMS/HW_FREE from one thread while another thread mmaps the same PCM stream, and must win a narrow timing window (assessed CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H); no public exploit code has been identified at time of analysis, although the upstream commit states a reproducer is available on request, and this CVE is not in CISA KEV. Vendor patches are available in the stable trees (6.12.111, 6.18.53, 7.2.7) and mainline 7.3-rc2, with backports to older long-term branches, so unpatched kernels running ALSA PCM support remain exposed to any local user who can open a PCM device.
A use-after-free race condition in the Linux kernel's ksmbd in-kernel SMB3 server can crash the kernel (denial of service) when a session is torn down while multiple channels are bound to a session and deferred byte-range locks or pending CANCELs remain counted as running requests. Affected systems must have ksmbd enabled (CONFIG_SMB_SERVER) and exporting at least one SMB3 share, and an authenticated remote attacker (PR:L) must negotiate SMB3 multichannel to race a LOGOFF, synchronous CANCEL, or abrupt connection teardown against still-running requests; success is non-deterministic (AC:H). No public exploit code has been identified at time of analysis, and the impact is availability loss, not information disclosure or confidentiality/integrity compromise as some tags or the base CVSS may imply.
Use-after-free race in the Linux kernel's ksmbd SMB3 server can be triggered when a network-device event such as an interface up/down operation invokes ksmbd's netdevice notifier during listener teardown, potentially allowing a local actor who can induce netdev events to corrupt memory and cause denial of service. Affected systems are those with CONFIG_SMB_SERVER enabled and at least one ksmbd listener bound to a network interface; hosts using userspace Samba instead of ksmbd are not affected. No CISA KEV entry or public exploit code was provided, and no public exploit identified at time of analysis; the assessed vector is local, high complexity, PR:L, C:N/I:L/A:H, so the NVD 7.8 vector overstates confidentiality and integrity risk, while vendor-released fixes are available in Linux 7.2.7 and 7.3-rc2.
A stack out-of-bounds write in the Linux kernel's Bluetooth L2CAP LE connection path lets an attacker within Bluetooth LE radio range corrupt kernel memory by sending L2CAP_LE_CONN_REQ frames - including duplicate identifiers - to a host running a userspace Bluetooth application that listens on an L2CAP LE channel with L2CAP_MODE_EXT_FLOWCTL and FLAG_DEFER_SETUP set. The CVSS base score is 7.5, but the independent assessment rates actual impact as integrity and availability only (no confidentiality impact), gated on that specific, non-default socket configuration, physical radio proximity, and affected kernels from 5.7 onward; no public exploit code has been identified and no active exploitation is confirmed. Unauthenticated (PR:N) adjacent attackers are the relevant threat model per the CVSS vector, and vendor patches are available.
Out-of-bounds memory access in the Linux kernel's IPv6 Segment Routing (SRv6) receive path can be triggered by remote, unauthenticated attackers who send a crafted IPv6 packet containing a Segment Routing Header (SRH) with segments_left > 0 that is preceded by another extension header such as a Hop-by-Hop options header. Because ipv6_srh_rcv() restores skb->data by pushing a fixed sizeof(struct ipv6hdr) instead of re-establishing the real network header, skb_network_offset() remains negative; that negative value is then consumed as an unsigned length by the flow dissector during ip6_route_input(), or by sch_fragment()/neighbour output when the packet is forwarded via ip6_forward() or redirected via act_mirred, producing an out-of-bounds memcpy or buffer overflow that can crash the kernel and corrupt memory. Exploitation is not automatic: the target must be built with CONFIG_IPV6_SEG6 and configured as an SRv6 endpoint, and the packet must be specifically structured, which is reflected in the assessed vector AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:H; no public exploit code and no confirmed active exploitation were identified at time of analysis. Vendor fixes are available in Linux 6.12.111, 6.18.53, 7.2.7 and 7.3-rc2.
A use-after-free in the Linux kernel's btrfs RAID stripe-tree insertion path can free an in-flight transaction handle and transaction twice, corrupting kernel memory on systems that enable the opt-in raid-stripe-tree feature (added in Linux 6.7). Only local users who can write to such a filesystem, and only when a memory allocation fails (-ENOMEM) at exactly the btrfs_insert_one_raid_extent() failure site during ordered-extent finishing, can trigger it; the independent assessment scores this AV:L/AC:H/PR:L with no confidentiality impact but high availability impact (C:N/I:L/A:H), noting it is a high-CVSS-but-low-real-risk kernel issue that is not attacker-input-controlled. There is no public exploit identified at time of analysis and the flaw is not confirmed actively exploited (CISA KEV); vendor-released patches exist in the affected branches.
A local altname-conflict handling flaw in the Linux kernel's network namespace teardown path (__dev_change_net_namespace()) allows an attacker with CAP_NET_ADMIN in a network namespace to trigger a WARN_ON assertion splat in cfg80211_pernet_exit() by assigning the same alternative interface name to devices in both the root netns and a non-root netns and then destroying the non-root netns. The affected code is the wireless (cfg80211/cfg802154) per-netns cleanup, which previously lacked the conflict-removal logic already added for default_device_exit_net(); the fix generalizes that cleanup into __dev_change_net_namespace(). Exploitation requires local access and the ability to create network namespaces and devices - achievable by an unprivileged user where unprivileged user namespaces are enabled - but the practical impact is limited to a kernel warning on default configurations, escalating to a real denial of service only when panic_on_warn=1 is set; no public exploit code has been identified at time of analysis, and this issue is not confirmed actively exploited (CISA KEV). Our independent assessment rates this as low real risk and considers the supplied CVSS 7.8 (C:H/I:H/A:H) materially overstated.
A race condition in the Linux kernel's RDS-over-TCP transport (net/rds) lets a local user who can open and use an AF_RDS socket corrupt connection state when a transmit is in flight at the same moment the accept path resets the connection's callbacks. Because rds_tcp_reset_callbacks() only samples the RDS_IN_XMIT bit as clear rather than owning it, rds_send_xmit() can re-take the bit immediately afterwards and run concurrently with rds_send_path_reset() rewriting cp_xmit_* state, producing torn transmit state and a potential kernel crash/denial of service with limited integrity impact. The authoritative assessment scores this CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:H - local access with low privileges, a narrow timing window, and, critically, manifestation restricted to weakly ordered CPUs (ARM, PowerPC) since the store-buffering pattern cannot occur on strongly ordered x86; no public exploit code has been identified at time of analysis and there is no confirmation of active exploitation (no CISA KEV entry), although the public tag of 'Information Disclosure' is a mischaracterization of what is really a CWE-362 concurrency bug. Vendor-released patches exist for the stable branches (6.12.111, 6.18.53, 7.2.7, and 7.3-rc2).
The Linux kernel's Reliable Datagram Sockets (RDS) connection teardown path contains a race condition in rds_conn_shutdown() that can crash or hang the kernel, causing denial of service on systems where the RDS subsystem is loaded and connections are being shut down or failed over. The flaw stems from teardown merely sampling the RDS_IN_XMIT and RDS_RECV_REFILL fastpath bits as clear instead of owning them, so a concurrent rds_send_xmit() or rds_ib_recv_refill() can re-acquire the bit and run while the transport re-initializes its rings and rds_send_path_reset() rewrites transmit state under it. Because the store-buffering pattern only manifests where memory ordering is weak (for example ARM), and because the accept-side state check must pass before the path moves to RDS_CONN_DISCONNECTING, reliable triggering is difficult and the practical impact is limited to kernel instability; no public exploit code has been identified at time of analysis and vendor patches are available.
A soft-lockup denial of service in the Linux kernel NVMe driver lets an attacker-controlled or misbehaving NVMe target stall host CPUs by advertising a sparse 32-bit namespace ID list containing a large gap, because nvme_scan_ns_list() walks every missing NSID one at a time instead of only the namespaces actually present. Practical exploitation is not drive-by: the host administrator must first attach the host to the hostile target, either over NVMe-over-Fabrics (an 'nvme connect' style operation) or through a crafted local NVMe device, and the resulting impact is availability only (C:N/I:N, A:H; assessed vector CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H, feed score 7.5). No public exploit identified at time of analysis and no CISA KEV entry is associated with this CVE, but vendor-released patches exist in the 6.12.111, 6.18.53, 7.2.7 and 7.3-rc2 kernel releases.
Use-after-free in the Linux kernel's bcmasp driver for Broadcom ASP 2.0 Ethernet hardware can cause premature release of an sk_buff when a reused TX descriptor retains a stale "last" flag, affecting only systems with that specific Broadcom NIC/SoC and requiring local access plus fragmented transmission activity that leads to descriptor reuse. The assessed impact is primarily availability (A:H) with limited integrity (I:L) and no confidentiality loss (C:N) under CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:H, and no public exploit identified at time of analysis. The original base score of 7.8 and C:H/I:H/A:H overstate real-world risk because the trigger requires specific hardware, local privileges, and a non-deterministic timing-dependent TX-ring condition.
A concurrency flaw in the Linux kernel's mlxsw Spectrum PTP driver (spectrum_ptp) can panic the kernel when a stale PTP entry is reaped by the garbage-collection workqueue at the same moment the threaded NAPI poll is running on another CPU. The GC path calls mlxsw_sp1_ptp_unmatched_finish() → mlxsw_sp1_ptp_packet_finish() → mlxsw_sp_rx_listener_no_mark_func(), which ends in napi_gro_receive() using a NAPI pointer captured earlier in the RX context; because mlxsw_pci_napi_devs_init() unconditionally calls netif_threaded_enable(), that poll runs as an independently scheduled kthread that local_bh_disable() on the GC CPU cannot protect against, so both CPUs mutate the same GRO list and hit list_add corruption / 'kernel BUG at lib/list_debug.c'. Exploitation is not remotely trivial: it requires physical Mellanox Spectrum hardware running the mlxsw driver, PTP hardware timestamping configured and active (e.g. ptp4l) with privileged CAP_NET_ADMIN access, plus a genuine timing race (AC:H) - the reported trigger is a tx-timestamp timeout while a port transitions to UNCALIBRATED/SLAVE. No public exploit code or in-the-wild activity has been identified at time of analysis, and our independent assessment rates the published CVSS 7.5 (AV:A/PR:N) as an overstatement, scoring the issue as a locally triggered, privilege-gated denial-of-service/kernel-panic (CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:H).