Buffer Overflow

Memory disclosure in the Windows IKE (Internet Key Exchange) Extension lets an attacker who already holds low-level authorized access on the target read out-of-bounds process memory and leak potentially sensitive data over the network. The CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N, 6.5) confirms a low-complexity, network-reachable attack that requires existing privileges, needs no user interaction, and produces only confidentiality impact with no integrity or availability effect. Microsoft has released vendor patches, no public exploit code has been identified at time of analysis, and CISA's SSVC assessment records exploitation as 'none' with partial technical impact, so this should be treated as a normal-cycle patch rather than an emergency.

Out-of-bounds read in the Windows Win32 kernel subsystem (win32k) lets an attacker who already holds a low-privileged local account on a Windows host read kernel memory it should not be able to access, resulting in information disclosure only (CVSS 3.1 5.5, AV:L/PR:L/UI:N, C:H/I:N/A:N). Affected platforms span Windows 10 (1607 through 22H2), Windows 11 (23H2, 24H2, 25H2, 26H1, 26H2) and Windows Server 2012, 2012 R2, 2016, 2019, 2022 and 2025, including Server Core installations. There is no public exploit identified at time of analysis and CISA's SSVC assessment records exploitation as none with no automation, while Microsoft has shipped a vendor patch; the practical risk is bounded by the requirement for an existing foothold, since the flaw cannot be reached remotely or without local code execution.

Out-of-bounds read in the Windows Win32 Kernel Subsystem lets a locally authenticated attacker holding low-privilege execution rights read beyond the bounds of a kernel buffer and disclose memory contents on affected Windows 10, Windows 11, and Windows Server builds; no user interaction is required. Scoring 5.5 (Medium) under CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N, the flaw is an information-leak primitive only - it provides no direct code execution or privilege escalation, and its value depends on being chained with a separate escalation or KASLR-bypass technique. There is no public exploit identified at time of analysis, and Microsoft has shipped fixes for all listed build families via the MSRC update guide.

Local information disclosure in the Windows Win32 kernel subsystem (win32k) allows an attacker who already holds low-privileged, authenticated local access to read protected kernel memory and potentially disclose sensitive data. Affected builds span Windows 10 (1607, 1809, 21H2, 22H2), Windows 11 (23H2, 24H2, 25H2, 26H1, 26H2), and Windows Server 2012 through 2025, including Server Core installations. Microsoft has released fixes through its update guide, and there is no public exploit identified at time of analysis; the vulnerability is not listed in CISA KEV and CISA's SSVC assessment rates exploitation as none and automation as no, so real-world urgency is LOW and the flaw is confined to post-compromise scenarios rather than initial access.

An out-of-bounds read in the Windows Fast FAT driver (fastfat.sys) lets a local attacker disclose kernel memory contents by persuading a user to mount or open an attacker-crafted FAT/exFAT volume, disk image, or VHD. Windows 10 and Windows 11 (23H2 through 26H2) plus Windows Server 2012 through 2025 are affected, and exploitation requires local access with user interaction while yielding information disclosure only - no code execution or system modification, which keeps severity moderate rather than critical. No public exploit code has been identified at time of analysis, no confirmed active exploitation is known, and a vendor patch is available from Microsoft's update guide.

Out-of-bounds read in Microsoft Word lets a remote, unauthenticated attacker disclose memory contents from an affected host when a victim opens or previews a specially crafted Word document, producing confidentiality-only impact (C:H/I:N/A:N) under a 6.5 CVSS score. Affected builds include Office LTSC 2021 before 16.0.14334.20974, Office LTSC 2024 before 16.0.17932.21056, Microsoft 365 Apps for Enterprise, Office 365 for Mac, and Office LTSC for Mac 2021/2024; the required user interaction (UI:R) is the main limiting factor, since nothing triggers the parser automatically. No public exploit code has been identified at time of analysis, the CVE is not in CISA KEV, and CISA SSVC records Exploitation as none and Automatable as no, so the risk signals are consistently moderate and non-urgent.

An out-of-bounds read (CWE-125) in the Windows DNS Server role lets an unauthenticated remote attacker disclose process memory contents over the network; per the CVSS vector (AV:N/PR:N/UI:R) exploitation is not fully autonomous, because the server must be induced to parse attacker-influenced DNS data - for example by resolving an attacker-controlled name or handling a crafted response - rather than being triggered by a single unsolicited packet. Impact is limited to confidentiality (C:H/I:N/A:N): the flaw yields a memory over-read and information disclosure, not remote code execution or service disruption, and the practical risk is highest for internet-exposed or recursive/forwarding DNS servers. No public exploit code or confirmed active exploitation has been identified at time of analysis, and the vendor has released patched builds for all listed Windows client and server versions, making this a moderate-priority information-disclosure issue that is removed by applying the Microsoft update.

Local information disclosure in the Windows Desktop Window Manager (DWM) lets an attacker who already holds low-privileged code execution on a host trigger a buffer over-read and harvest out-of-bounds memory contents from the desktop compositor that is present and running by default on every affected system. The issue is reachable only with local access at low privilege and requires no user interaction, so it cannot be leveraged remotely or used on its own to gain code execution, escalate privileges, or cause a denial of service. No public exploit code has been identified at time of analysis and there is no confirmed active exploitation, and because the impact is limited to confidentiality (CVSS 5.5, C:H/I:N/A:N), this is assessed as a medium-severity, low-priority issue despite the ubiquity of the affected product; all supported Windows 10, Windows 11, and Windows Server builds below the fixed build numbers in Microsoft's update guide are affected.

Local information disclosure in the Windows Device Association Service (the component that handles device pairing and association) allows an attacker who already holds low-privileged, authenticated access on a Windows host to read out-of-bounds memory and potentially expose sensitive data. Microsoft rates this CVSS 3.1 5.5 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N), reflecting a local-only vector with no user interaction, no integrity or availability impact, and unchanged scope. There is no public exploit identified at time of analysis and CISA's SSVC assessment records no exploitation activity, so this is a moderate, low-priority issue best handled through routine patching rather than emergency response.

An out-of-bounds read in Microsoft Excel lets remote attackers disclose information when a victim opens a specially crafted spreadsheet, with no authentication or prior privileges required (CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N). Exploitation is not zero-click: the attacker must deliver a malicious .xls/.xlsx file (email or web download) and convince the user to open it in a vulnerable build, and there is no public exploit code and no confirmed active exploitation (SSVC Exploitation: none, Automatable: no; no CISA KEV listing) at time of analysis. Affected builds are Microsoft Office LTSC 2021 before 16.0.14334.20974, Microsoft Office LTSC 2024 before 16.0.17932.21056, and Microsoft 365 Apps for Enterprise, all of which have vendor patches available; impact is limited to confidentiality, with no integrity or availability effect.

Out-of-bounds read in Microsoft Excel allows local attackers to disclose sensitive information from memory when a user opens a crafted spreadsheet. Affected builds include Microsoft 365 Apps for Enterprise, Office LTSC 2024 (below 16.0.17932.21056), and Office LTSC 2021 (below 16.0.14334.20974). Exploitation requires the victim to open a malicious file locally (UI:R, AV:L, PR:N), no public exploit was identified at time of analysis (SSVC Exploitation=none, Automatable=no), and vendor patches are available; the risk is assessed as moderate with impact limited to confidentiality (C:H, I:N, A:N) and no code execution or data modification.

Information disclosure in Windows Win32k allows an authenticated local attacker with low privileges to trigger an out-of-bounds read in the kernel-mode Win32 subsystem, potentially exposing sensitive kernel memory contents. The CVSS 5.5 rating reflects a local-only attack path (AV:L/PR:L) with no user interaction and no direct integrity or availability impact, so the flaw cannot be used for initial access and is most relevant as an info-leak primitive within a larger local exploit chain. Vendor patches are available, and no public exploit code or active exploitation has been identified at time of analysis.

An out-of-bounds read in the Windows HTTP.sys kernel-mode HTTP stack lets an authenticated local user force the driver to read beyond the bounds of an allocated buffer and disclose sensitive memory contents, rated CVSS 5.5 (AV:L/PR:L/UI:N/C:H/I:N/A:N). Only hosts that already grant an attacker a local foothold with a valid low-privileged account are exposed; remote and unauthenticated attackers are excluded, and no user interaction or special complexity is required once that access exists. There is no confirmed active exploitation (CISA KEV absent) and no public exploit identified at time of analysis, and Microsoft has released fixed builds for all affected Windows client and server versions, so patched systems are not exploitable.

Out-of-bounds read in Microsoft Office Word allows a remote, unauthenticated attacker to disclose the contents of memory beyond the bounds of an allocated buffer when a victim opens a specially crafted document (CVSS 6.5, AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N). Affected builds include Office LTSC 2021 and 2024 for Windows, Microsoft 365 Apps for Enterprise, and the Office LTSC/365 editions for Mac, so the exposure footprint is essentially the entire supported Microsoft Office desktop install base. Impact is confidentiality-only - read access to process memory, with no integrity or availability effect and no demonstrated code execution path - and the flaw cannot be triggered without the user opening the malicious file. No public exploit code or active exploitation has been identified at time of analysis, and CISA's SSVC assessment rates exploitation as none and automatable as no, keeping real-world urgency low despite the ubiquitous target.

Local denial of service in Windows DHCP Server (CVE-2026-70201) allows an authenticated attacker holding valid low-privilege credentials on a host running the DHCP Server role to crash the service by supplying malformed input that triggers an out-of-bounds read (CWE-125). Per the assessed CVSS 3.1 vector (AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H, score 5.5) the issue is not remotely reachable and yields only availability impact, with no confidentiality or integrity loss and no code execution path. No public exploit identified at time of analysis: CISA SSVC records exploitation as 'none', automatable as 'no', technical impact as 'partial', and the CVE is not KEV-listed; Microsoft has shipped vendor patches, so urgency is low and the main prerequisite is that the DHCP Server role be installed and running (it is not enabled by default on typical member servers).

An out-of-bounds read in the Microsoft Windows Search Component lets an authenticated attacker with only low-privileged local access cause the component to read past the end of a buffer and disclose information from process memory. Exploitation requires the attacker to already hold local interactive access to the host (CVSS:3.1/AV:L/AC:L/PR:L/UI:N) and to induce the Search Component to process attacker-influenced input; there is no network or unauthenticated path, and the outcome is memory disclosure only, with no integrity or availability impact. No public exploit code has been identified and CISA's SSVC assessment records exploitation as none, so this is a moderate-severity (CVSS 5.5) local information-disclosure issue with low real-world urgency despite its high confidentiality rating, though the vendor has released fixed builds for all affected Windows editions.

An out-of-bounds read in Windows DNS Server allows an authenticated, low-privileged attacker to cause a denial of service (DoS) over the network. The flaw affects Windows Server 2012, 2012 R2, 2016, 2019, 2022, 2025 and Windows 10 Versions 1607 and 1809 when the DNS Server role is enabled; exploitation requires valid credentials (PR:L) but no user interaction. No public exploit code has been identified, and CISA SSVC rates exploitation as none, yielding a modest, non-urgent risk despite the network attack vector.

An out-of-bounds read in the Microsoft Windows Search component allows an attacker who is already authenticated on the local machine (CVSS AV:L/PR:L) to disclose information from process memory, with no code execution, integrity, or availability impact (C:H/I:N/A:N) for a base score of 5.5 (MEDIUM). Affected systems are unpatched Windows client and Server builds, including Windows 10 21H2/22H2, Windows 11 23H2 through 26H2, and Windows Server 2012 through 2025; systems upgraded to the fixed builds (for example Windows 11 24H2 10.0.26100.9711 or Windows Server 2025 10.0.26100.33650) are not affected. No public exploit code has been identified at time of analysis, CISA's SSVC assessment rates exploitation as none and automation as not feasible, and the flaw is not listed in CISA KEV, so this is a low-priority, moderate-severity issue best handled through normal patch cycles rather than emergency response.

Information disclosure from an out-of-bounds read in the Windows Remote POS Driver allows an authenticated local user to read memory past the boundary of an intended buffer, potentially exposing sensitive data handled in kernel context. Exploitation requires local access with low privileges (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N), needs no user interaction, and is limited to confidentiality impact - no integrity or availability effect - which caps real-world severity despite the HIGH confidentiality rating. No public exploit code and no confirmed active exploitation were identified at time of analysis (CISA SSVC: Exploitation = none, Automatable = no), and the requirement for pre-existing local authenticated access is the principal factor reducing risk.

An out-of-bounds read in the HID Parser Library shipped with Microsoft Windows allows an attacker who already holds local access and low-level privileges on the host to disclose information from kernel memory. The flaw is reachable when the parser processes a malformed HID report descriptor, which can come from an attacker-controlled USB or Bluetooth HID device or from a local process feeding crafted HID data; CVSS 3.1 scores it 5.5 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N), meaning confidentiality impact only with no integrity or availability effect. No public exploit code or active exploitation has been identified at time of analysis (CISA SSVC exploitation: none, automatable: no), and Microsoft has published a vendor patch through the MSRC update guide, so urgency is moderate and local in scope.

An out-of-bounds read in the Windows DNS Server service on affected Windows Server builds permits an authenticated, network-positioned attacker to trigger memory access beyond an allocated buffer, with the vendor's own vector scoring the impact as availability-only (A:H) rather than code execution. Exploitation requires network reachability to a DNS service, low attack complexity, no user interaction, and low-level authorization (PR:L), meaning the attacker must already hold valid credentials or an authorized position rather than being fully anonymous; the vendor-derived description claims possible code execution, but that claim is not supported by the CVSS impact metrics and should be confirmed with Microsoft. No public exploit code has been identified at time of analysis, the CISA SSVC assessment records no known exploitation activity, and the defect is not listed in KEV, so the realistic near-term risk is disruption of DNS resolution on unpatched Windows Server 2022 and Windows Server 2025 hosts.

Stack-based buffer overflow in the Windows Resilient File System (ReFS) driver allows an unauthenticated attacker with direct physical access to a device - for example by attaching or mounting an attacker-controlled ReFS-formatted volume - to corrupt kernel memory and bypass a security feature, resulting in a high integrity impact with no loss of confidentiality or availability. The assessed vector (AV:P/AC:H/PR:N/UI:N/S:U/C:N/I:H/A:N) makes remote or scaled exploitation infeasible: the attacker must be physically present and the overflow is not reliably triggerable, and CISA's SSVC evaluation records no exploitation activity and no automatable path. No public exploit code has been identified at time of analysis.

An out-of-bounds read in the Windows DNS component lets a locally authenticated, low-privileged attacker cause the service to read past the end of an allocated buffer and leak information from process memory, giving high confidentiality impact with no integrity or availability effect (CVSS 3.1 base 5.5, AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N). Exploitation requires an attacker to already hold local, authenticated low-privileged access to an affected Windows host - there is no remote or network-triggerable path - and no public exploit code has been identified at time of analysis, with CISA SSVC recording exploitation as 'none' and automatable as 'no'. Affected builds span Windows Server 2012 through Windows Server 2025 and Windows 10 versions 1607 and 1809, and Microsoft has released fixed builds for each, making this a moderate, low-priority patching item rather than an urgent emergency.

Information disclosure in the Windows Kernel lets an attacker who already holds local, authenticated low-privilege access on a Windows host read past the bounds of a kernel buffer and leak adjacent kernel memory, but only when a hard-to-control precondition such as a race window or a specific kernel memory layout is satisfied. All currently supported Windows 10, Windows 11 and Windows Server builds below the patched levels listed in Microsoft's advisory are affected; the defect can neither modify data nor crash the system, so its impact is confined to confidentiality (C:H/I:N/A:N, CVSS 4.7). No public exploit code has been identified at time of analysis, CISA's SSVC assessment records exploitation as 'none' and automatable as 'no', and no EPSS score was supplied with this record - the limiting factors are the local-access prerequisite and the complex triggering condition, making this a low-to-moderate priority despite the notable 'Windows Kernel' target.

An out-of-bounds read (CWE-125) in Microsoft's Windows Media Foundation allows a local attacker to disclose information by convincing a victim to open or process a specially crafted media file. Affected are unpatched Windows 10, Windows 11, and Windows Server editions up to the fixed build for each release - for example Windows 11 24H2 below 10.0.26100.9711, Windows 10 22H2 below 10.0.19044.7807, Windows Server 2022 below 10.0.20348.5761, and Windows Server 2012 R2 below 6.3.9600.23470. Exploitation requires no attacker privileges (PR:N) but is entirely contingent on victim user interaction (UI:R) with the malicious file, and there is no public exploit identified at time of analysis; CISA's SSVC assessment rates exploitation as none and automatable as no, and the impact is read-only information disclosure with no code execution or data tampering.

Information disclosure in the Windows CD-ROM driver (cdrom.sys) lets an attacker who already holds a low-privileged local session on a Windows host read past the bounds of an allocated buffer in kernel memory, exposing data they should not be able to see. Per the assessed vector (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N, 5.5), the flaw is local-only, requires prior authenticated access, needs no victim interaction, and is confined to confidentiality impact - it cannot alter data or crash the system, so remote and unauthenticated attackers cannot use it directly. No public exploit code and no confirmed active exploitation were identified at time of analysis, and Microsoft has released fixed builds across Windows 10/11 and Windows Server 2012 through 2025, making this a moderate, low-priority item despite the ubiquity of the affected platform.

An out-of-bounds read in the Windows IP Over USB Service (the component that exposes IP connectivity to devices over USB, typically used by device and SDK tooling rather than being a default always-on service) lets an attacker who already holds low-privileged local credentials read beyond intended memory boundaries and disclose information. According to the assessed CVSS 3.1 vector AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N, this is a local, authenticated, no-user-interaction information disclosure with high confidentiality impact but no integrity or availability impact, and there is no remote vector, privilege escalation, or code execution. There is no public exploit identified at time of analysis and CISA's SSVC assessment rates exploitation as none with the issue not automatable; this is a moderate-severity, low-priority issue rather than an urgent one, and vendor patches are available.

Information disclosure in the Microsoft Windows Update component allows a local, low-privileged authenticated attacker (CVSS AV:L/PR:L, UI:N) to trigger an out-of-bounds read and expose contents of process memory. Only specific Windows 11 builds (23H2, 24H2, 25H2, 26H1, 26H2) and Windows Server 2025 - including Server Core - below the fixed build numbers are affected; the flaw cannot be reached remotely or by an unauthenticated actor, and impact is limited to confidentiality (C:H) with no integrity or availability effect and no code execution or privilege escalation on its own. No public exploit code has been identified at time of analysis, CISA SSVC records exploitation as 'none', and vendor patches are available, making this a moderate, low-priority issue when all signals are weighed together.

Information disclosure in the Windows Desktop Window Manager (dwm.exe) allows a locally authenticated attacker holding only low-level user privileges to read memory beyond an intended buffer inside the compositor process, potentially exposing data belonging to the DWM session. Exploitation is not straightforward: it requires an existing local foothold and the high attack complexity indicated by the CVSS vector (AV:L/AC:H/PR:L/UI:N) means success likely depends on winning a race or hitting a specific memory state that the attacker cannot reliably control. No public exploit code has been identified at time of analysis, CISA's SSVC assessment records exploitation as 'none' (Automatable: no), and the issue is confined to confidentiality impact (C:H/I:N/A:N); Microsoft has released fixes for all affected Windows branches, making this a low-priority local information-disclosure flaw that is best addressed through routine patch cycles.

Local information disclosure in the Windows Overlay Filter (WOF) component of supported Windows client and server releases allows an authenticated attacker who already holds a foothold on the host to read memory beyond an allocated buffer, potentially leaking kernel or process memory contents. Exploitation requires local access as a low-privileged user (CVSS:3.1/AV:L/PR:L/UI:N) and the ability to exercise the WOF code path used for file-system compression and overlay operations; the outcome is confidentiality-only (C:H, I:N, A:N) with no code execution, integrity, or availability impact. No public exploit code or active exploitation has been identified at time of analysis (SSVC exploitation: none, automatable: no), and a vendor patch is available, so the practical risk is that of a moderate information-leak primitive that could be chained into a broader exploit rather than a standalone compromise.

Code execution with full confidentiality, integrity and availability impact is achievable on unpatched Windows systems through a heap-based buffer overflow in the Windows Overlay Filter (WOF), but only for an attacker who already holds physical access to the target machine. The flaw carries a base score of 6.8 (CVSS:3.1/AV:P/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H): physical access is required, yet once obtained no authentication and no user interaction are needed and attack complexity is low, so any unattended, shared or physically exposed endpoint that processes attacker-supplied input through WOF is at risk. No public exploit code has been identified at time of analysis and CISA's SSVC decision framework rates exploitation as none and automation as not feasible, but Microsoft has published a vendor fix and the technical impact is assessed as total, so the residual risk is concentrated in kiosks, lobby/plant-floor PCs, and other hands-on devices rather than in remotely reachable server fleets.

An out-of-bounds read in the Windows Client for NFS (Services for NFS) allows an unauthenticated, network-positioned attacker to leak client memory contents when a victim connects to or mounts a share from an attacker-controlled NFS server that returns crafted protocol responses. Exploitation is narrow: the non-default optional 'Client for NFS' feature must be installed, the victim must initiate the connection (UI:R), and the impact is confidentiality-only with no code execution or data modification. There is no public exploit code and no confirmed active exploitation (SSVC Exploitation: none; not listed in CISA KEV), and Microsoft has shipped vendor updates, making this a moderate-severity information-disclosure issue to remediate on a normal patch cadence rather than an emergency.

An authenticated local attacker can trigger an out-of-bounds read (CWE-126) in the Windows Overlay Filter (WCIFS) driver and disclose the contents of adjacent kernel memory on affected Windows 10, Windows 11 and Windows Server builds. The flaw is only reachable from an existing interactive foothold on the host with at least low user privileges (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N, base 5.5) and results in confidentiality loss only, with no integrity or availability effect. There is no public exploit identified at time of analysis, CISA's SSVC rates exploitation as 'none' and the issue as non-automatable, and Microsoft has shipped fixed builds for all listed products.

Code execution in the Windows Overlay Filter (WOF) driver of unpatched Windows 10 (21H2/22H2), Windows 11 (23H2 through 26H2), and Windows Server 2022/2025 allows an unauthenticated attacker to gain full control of confidentiality, integrity, and availability on the affected host, but only if they already hold physical, hands-on access to the device (CVSS AV:P), which caps the base score at 6.8. The root cause is a heap-based buffer overflow (CWE-122) that is reachable with low attack complexity, no privileges (PR:N), and no user interaction (UI:N), so the decisive limiting factor is the physical-access prerequisite rather than any technical barrier. No public exploit code and no confirmed active exploitation were identified at time of analysis (CISA SSVC Exploitation=none, Automatable=no), and a vendor patch is available, so remote, wormable, or mass-scale abuse is not feasible.

An out-of-bounds read in the Windows Client for NFS (the optional Services for NFS client component) allows a remote attacker to leak contents of process memory from an affected Windows host, yielding information disclosure only with no code execution, integrity, or availability impact. Because the feature is not present in a default Windows installation, the exposed population is limited to hosts where the NFS client has been explicitly installed, and per the assessed vector (CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N, 6.5) exploitation also requires user interaction - the victim must initiate or be induced into connecting to or mounting an attacker-controlled or spoofed NFS resource, though no authentication is required on the target. No public exploit code has been identified at time of analysis, CISA's SSVC assessment rates exploitation as 'none' with automatable 'no' and technical impact 'partial', and Microsoft has released a vendor patch.

An out-of-bounds read in Microsoft Excel enables an unauthenticated attacker to disclose sensitive in-process memory by convincing a victim to open a specially crafted Excel document (CVSS 3.1: 6.5, AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N). Affected builds are Microsoft 365 Apps for Enterprise, Microsoft Office LTSC 2021 before 16.0.14334.20878, and Microsoft Office LTSC 2024 before 16.0.17932.20934; the mandatory user interaction (UI:R) is the primary limiter and prevents automated or wormable exploitation, matching the SSVC assessment of Exploitation: none and Automatable: no. There is no public exploit identified at time of analysis, and the impact is limited to information disclosure - no code execution, data modification, or denial of service.

Local information disclosure in Windows Cryptographic Services (CryptSvc) allows an attacker who already holds authenticated, low-privileged access to the host to trigger an out-of-bounds read and leak bytes from process memory, potentially including cryptographic material handled by the service. Only Windows Server 2025 (including Server Core installations) and Windows 11 24H2, 25H2, 26H1, and 26H2 builds older than the vendor-fixed build numbers are affected; Microsoft has released patches, and no public exploit identified at time of analysis. The CVSS 3.1 score of 5.5 (AV:L/PR:L/UI:N, high confidentiality impact only) alongside a CISA SSVC rating of no exploitation and not automatable places real-world urgency at a low level.

An out-of-bounds read (CWE-125) in the Windows AppX Deployment Service allows an attacker who already holds an authenticated, low-privileged local session on a Windows host to read past the bounds of a memory buffer and disclose information to which they should not have access. The issue is rated CVSS 5.5 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N), so it requires local access and valid credentials but no user interaction, and it affects only confidentiality - there is no integrity or availability impact and no privilege escalation is described. Exploitation has not been observed (CISA SSVC exploitation: none) and there is no public exploit identified at time of analysis; vendor patches are available, so this is a moderate-severity, low-urgency item that should be resolved through routine monthly patching rather than emergency action.

An out-of-bounds read (CWE-125) in the Windows iSCSI component allows a locally authenticated, low-privileged attacker to disclose sensitive memory contents from the affected host, rated CVSS 5.5 (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N). Exploitation is strictly local: the attacker must already hold valid low-privileged credentials on the machine and reach the vulnerable iSCSI code path - no remote or unauthenticated vector and no victim interaction are involved, and the impact is confidentiality-only, with no code execution, integrity, or availability loss on its own. Vendor-released patches are available through Microsoft's MSRC update guide, and no public exploit code or confirmed active exploitation (CISA KEV) has been identified at time of analysis; the primary risk is as a memory-disclosure/ASLR-bypass primitive chained into a larger local attack rather than a standalone compromise.

Information disclosure in Windows NFS Server (Windows Network File System) allows a locally authenticated attacker with low privileges to trigger an out-of-bounds read and potentially expose sensitive memory contents on affected Windows Server hosts. Exploitation requires the non-default NFS server role to be installed and running, and the CVSS vector (AV:L/PR:L) confirms local access and low privileges; there is no confirmed active exploitation or public exploit code identified at time of analysis, and the issue is rated CVSS 5.5 (moderate). Microsoft has released patches, so affected organizations should treat this as a routine but non-urgent update for NFS-enabled servers.

Heap-based buffer overflow in the Windows Installer service lets an attacker who already holds an authenticated low-privilege session and has physical access to a vulnerable Windows machine escalate privileges, with high confidentiality, integrity, and availability impact within the operating system. The flaw is rated 6.6 (CVSS:3.1/AV:P/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) and affects a very broad install base - Windows 10 1607 through 22H2, Windows 11 23H2 through 26H2, and Windows Server 2012 through 2025 including Server Core - but it is not remotely reachable, requires no victim interaction, and is useless against patched builds. Vendor patches are available, no public exploit code or active exploitation has been identified at time of analysis, and CISA's SSVC assessment rates exploitation as 'none' and automatable as 'no', which makes this a moderate-severity, low-urgency local elevation-of-privilege issue rather than an urgent remote threat.

An out-of-bounds read (CWE-125) in the Windows Key Storage Provider (KSP) allows a locally authenticated attacker with high attack complexity to disclose potentially sensitive memory contents, though the impact is limited to confidentiality (CVSS 4.7, AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:N/A:N). All supported Windows releases from Server 2012 R2 through Server 2025 and Windows 10/11 are affected until patched to the fixed builds listed in the Microsoft advisory. No public exploit code or active exploitation (CISA KEV) has been identified at time of analysis. This is a low-to-moderate priority issue despite the memory-safety class.

Denial of service in the Windows Kernel can be triggered by an unauthenticated, network-based attacker who convinces a user to interact with crafted content, exploiting an out-of-bounds read (CWE-125). The vulnerability has a CVSS 3.1 score of 5.3 with availability-only impact (A:H; C:N/I:N); no data exposure or code execution is possible. No public exploit code has been identified at time of analysis, and CISA SSVC indicates no active exploitation, making real-world priority low. Microsoft has released patches.

An out-of-bounds read in the Windows DNS Server role lets a remote attacker who already holds low-level authorization (PR:L) read past the bounds of a buffer and disclose information over the network - confidentiality is fully impacted (C:H) but there is no integrity or availability effect, and no code execution. Affected platforms span the supported Windows Server line (2012, 2012 R2, 2016, 2019, 2022, 2025) plus Windows 10 (1607 through 22H2) and Windows 11 (23H2 through 26H2), so domain controllers and any host with the DNS Server role installed are the primary exposure. No public exploit has been identified at time of analysis and the issue is not confirmed actively exploited (not in CISA KEV); Microsoft has released a vendor patch, and this is a moderate-priority issue rather than an emergency.

Out-of-bounds read in the Microsoft Windows DNS Server role lets an attacker who already holds some level of authorization cause the service to read past the bounds of an allocated buffer and return that memory content over the network, disclosing information (CVSS 6.5, AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N). The flaw only matters where the DNS Server role is installed - standard on domain controllers and dedicated DNS infrastructure, but not enabled by default on client Windows - and per the PR:L vector the attacker must already be an authenticated/authorized principal, so this is not an unauthenticated remote bug. There is no public exploit identified at time of analysis, CISA SSVC records Exploitation=none with Automatable=no, and vendor patches are available for every affected build, so this is a moderate-priority information-disclosure issue rather than an emergency; note that although confidentiality impact is rated High, integrity and availability are untouched.

Information disclosure in Microsoft Windows DNS Server stems from an out-of-bounds read (CWE-125) that an authorized, network-positioned attacker can trigger to leak memory contents over the network. The CVSS:3.1 vector (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N, base 6.5) confirms remote reachability with low attack complexity, but PR:L indicates the attacker must already hold some authorized level of interaction with the DNS service rather than being fully anonymous; the impact is confined to confidentiality, with no integrity or availability effects, so DNS records cannot be altered and the service cannot be crashed through this flaw alone. Only hosts running the Windows DNS Server role are exposed - a server-side feature that is not present on default client installations - and no public exploit code has been identified at time of analysis (CISA SSVC exploitation: none, automatable: no, technical impact: partial); vendor patches are available.

Local information disclosure in Microsoft Defender for Endpoint for Windows arises from an out-of-bounds read that an authorized attacker with low-level privileges can trigger on a host to expose information from process memory. Only builds below 4.18.26060.3008 are affected, and the attack surface is strictly local (AV:L) with no user interaction required, so exploitation is confined to someone who has already gained a foothold on the target machine; the flaw yields only confidentiality impact (C:H) and cannot alter files, execute code, or crash the system (I:N/A:N). No public exploit code or confirmed active exploitation has been identified at time of analysis, and the 5.5 CVSS score places this in moderate, non-emergency territory.

Adobe FrameMaker 2026 (builds up to and including 18.0.0.206) and FrameMaker 2022 (builds up to and including 17.0.9.866) contain an out-of-bounds read that allows an attacker to disclose sensitive contents of process memory when a victim opens a specially crafted document. Exploitation is strictly local and depends on user interaction - the victim must open the malicious file - and the resulting impact is limited to information disclosure, with no code execution or system compromise (CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N, 5.5). No public exploit code has been identified at time of analysis and there is no indication of active exploitation, making this a low-urgency issue requiring remediation through the vendor advisory rather than emergency action.

Opening a crafted document in Adobe FrameMaker 2026 (versions up to and including 18.0.0.206) or Adobe FrameMaker 2022 (versions up to and including 17.0.9.866) can trigger an out-of-bounds read that discloses sensitive contents of the application's memory. The flaw requires no privileges or authentication from the attacker (PR:N) and is delivered as a local file that the victim must open (AV:L/UI:R), so exploitation depends on user interaction with attacker-controlled content rather than any network-reachable or service-side trigger; confidentiality impact is rated High while integrity and availability are unaffected (CVSS 5.5). No public exploit code has been identified at time of analysis and the issue is not listed as confirmed actively exploited (CISA KEV), so this is a moderate-severity, low-urgency information-disclosure flaw that fits normal patch-cycle urgency rather than emergency response.

Information disclosure in the Windows Broadcast DVR User Service lets a local, authenticated attacker trigger a CWE-126 buffer over-read and leak adjacent memory contents on affected Windows 10, Windows 11, and Windows Server 2019 hosts. Per the assessed vector (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N), the attacker must already hold low-privilege local access or code execution, no victim interaction is required, and impact is limited to confidentiality - there is no remote attack surface and no integrity or availability effect. There is no public exploit identified at time of analysis, and a vendor-released patch is available from Microsoft.

An out-of-bounds read in Microsoft Word lets an unauthorized attacker disclose information from process memory when a victim opens a crafted Word document, affecting Office LTSC 2021, Office LTSC 2024, and Microsoft 365 Apps for Enterprise builds that predate the vendor's fixed releases. Successful exploitation requires the user to open the malicious file delivered over the network (email attachment, web download, or file share); no attacker authentication is needed and default Office configurations are exposed. Impact is confined to confidentiality - process memory disclosure with no code execution, data modification, or availability effect - and no public exploit or confirmed active exploitation has been identified at time of analysis, making this a genuine but moderate-priority issue.

Heap-based buffer overflow in the Windows Portable Device (MTP) Driver lets an attacker who can physically attach a crafted MTP peripheral - a smartphone, camera, or rogue USB gadget - corrupt kernel heap memory and elevate privileges to SYSTEM on unpatched Windows 10, Windows 11, and Windows Server hosts. The CVSS vector (AV:P/AC:L/PR:N/UI:N) means no authentication and no user interaction are needed, but the physical-access prerequisite confines realistic risk to attended endpoints, kiosks, shared or public machines, and lost or stolen laptops rather than network-exposed systems. No public exploit code has been identified at time of analysis, and Microsoft has shipped patched builds, so this is a moderate-priority, patch-and-move-on issue rather than an emergency.

Denial of service in Microsoft's Windows Codecs Library affects default installations of Windows 10, Windows 11, and Windows Server 2012 through 2025 when an application invokes the vulnerable codec to parse a malformed media file. The defect is a stack-based buffer overflow (CWE-121) that is network-deliverable and requires no authentication (AV:N/PR:N), but it is gated by mandatory victim interaction (UI:R) such as opening, previewing, or rendering the crafted file, which prevents unattended remote exploitation. Impact is availability-only (A:H) with no confidentiality or integrity effect and no confirmed code execution; no public exploit code has been identified at time of analysis, the issue is not in CISA KEV, and Microsoft has released fixes. This is a moderate resilience/uptime concern rather than a breach vector.

Information disclosure in the Windows Remote Desktop Protocol lets an authenticated remote attacker trigger an out-of-bounds read and leak memory contents over the network, with affected builds spanning Windows 10 (1607 through 22H2), Windows 11 (23H2 through 26H2), and Windows Server 2012 through 2025. Exploitation requires the target's RDP service to be enabled and network-reachable plus valid credentials (PR:L, consistent with Microsoft's 'authorized attacker' wording), and a high-complexity precondition outside the attacker's control must also be satisfied (AC:H), which is why the CVSS score is held to 5.3 even though the confidentiality impact alone is rated High. Impact is confined to information disclosure - no integrity or availability effect - and no public exploit has been identified at time of analysis; Microsoft has published an update-guide entry and vendor patches are available.

Several buffer-growth sites in libmime performed size arithmetic in 32-bit types without overflow checks. On a sufficiently large message the growth guard could fail open, or the buffer could be reallocated smaller than required, and a subsequent memcpy would then write attacker-controlled data across adjacent heap allocations. This vulnerability was fixed in Thunderbird 158.

An out-of-bounds write in GPAC's MPEG-2 transport stream AIT parser (versions up to and including 26.07.0) allows remote attackers to corrupt memory by supplying a crafted stream whose AIT section declares more application descriptors than the fixed application_descriptors_id array can hold. Exploitation requires user interaction - someone or an automated workflow must feed the malicious stream into GPAC's M2TS/AIT parsing path - but no authentication or privileges are needed (CVSS:4.0 base 5.3; assessed CVSS:3.1 AV:N/AC:L/PR:N/UI:R, C:L/I:L/A:L). Publicly available exploit code exists (a PoC archive is linked from the VulDB entry and the upstream issue), though there is no evidence of confirmed active exploitation (not in CISA KEV); the practical risk is concentrated in services that automatically ingest untrusted media, such as transcoding pipelines and upload processors.

An out-of-bounds read in BlueZ's btmon Bluetooth monitoring utility, as packaged in Red Hat Enterprise Linux 6 through 10, allows an attacker to crash the tool by getting a user to open a crafted btsnoop capture file containing an HCI Return Link Keys event that advertises more link keys than the packet actually contains. The flaw is local-access and interaction-dependent (CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H) - the victim must deliberately load the malicious capture into btmon - which sharply limits real-world exposure given btmon is a diagnostic tool rarely pointed at untrusted input. No public exploit code or confirmed active exploitation was identified at time of analysis, and the practical outcome is a denial of service against the monitoring utility rather than any network- or privilege-escalation impact.

An out-of-bounds read in the ESP-IDF BlueDroid L2CAP configuration-option parser lets an unauthenticated Bluetooth BR/EDR peer within radio range send a malformed configuration request that reads past the end of the received packet, and under some runtime conditions limited adjacent Bluetooth stack memory can be echoed back in an outbound L2CAP Configuration Response. Only ESP32-class devices built with the BlueDroid host stack and Classic Bluetooth (BR/EDR) enabled are affected; builds using the NimBLE stack or with Classic Bluetooth disabled are not exposed. Assessed as CVSS 3.1 6.5 (Medium, AV:A/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:L), with no public exploit identified at time of analysis and no confirmed active exploitation - a genuine but moderate-priority issue rather than an emergency, addressed in 5.2.8, 5.3.6, 5.5.5, 6.0.3, and 6.1-rc1.

Local attackers with low-level privileges can cause buffer overflows and memory corruption in AMI AptioV UEFI firmware by supplying malformed length values to a firmware interface, potentially leading to arbitrary code execution within the BIOS/SMM environment. The vulnerability (CWE-805) affects AptioV-based firmware across OEM platforms and requires local access plus existing OS-level privileges; no public exploit identified at time of analysis. Due to the local attack vector and dependency on OEM firmware updates, this is assessed as a credible but second-tier priority.

An out-of-bounds write (CWE-787) in AMI Aptio V (APTIOV) BIOS firmware can be triggered by a local attacker who already holds high privileges, potentially corrupting memory inside firmware code paths and enabling buffer overflows, arbitrary code execution, and loss of confidentiality, integrity, and availability. Successful abuse requires local access plus pre-existing administrative or kernel-level privileges, and the high attack complexity means specific state, timing, or configuration preconditions must be met before the write can be controlled, so this is best characterized as a firmware persistence/privilege-escalation primitive rather than an initial-access or remote attack vector. At time of analysis there is no public exploit identified, no CISA KEV listing, and no confirmed exact fix version in the available data, though AMI has issued security advisory AMI-SA-2026002 covering the issue.

A classic buffer overflow (CWE-120) in AMI Aptio V UEFI/BIOS firmware allows a local user holding low-level privileges to supply crafted input to an affected BIOS code path, producing memory corruption that can crash or destabilize the platform (rated high availability impact) and potentially execute limited attacker-controlled code in firmware context. Exploitation is strictly local and requires existing privileges rather than anonymous access (CVSS:3.1/AV:L/AC:L/PR:L/UI:N), with no user interaction and low attack complexity, but the exact vulnerable interface is not identified because the CPE is wildcarded and the advisory content was not retrieved. There is no public exploit code identified at time of analysis and the flaw is not listed in CISA KEV, making this a serious-but-not-emergent local firmware flaw; no EPSS score was supplied.

Integer overflow in AMI AptioV BIOS firmware, classified as CWE-190, can be triggered by a local high-privileged caller to cause buffer overflows, memory corruption, and potential arbitrary code execution with high impact to confidentiality, integrity, and availability. Exploitation requires local access, an existing administrator/root-level foothold (PR:H), and high attack complexity (AC:H) per the assessed vector AV:L/AC:H/PR:H/UI:N/S:U/C:H/I:H/A:H, so it is not remotely reachable or mass-exploitable. No public exploit was identified at time of analysis, and active exploitation is not confirmed by KEV data; this is a moderate-priority, high-barrier firmware flaw rather than a widespread emergency.

FreeRDP's freerdp-proxy component (versions 3.17.1 up to but not including 3.32.0) can be induced to read heap memory out of bounds by a malicious or man-in-the-middle-controlled backend RDP server, which sends a DRDYNVC CREATE_REQUEST_PDU whose channel name is not NUL-terminated. The unbounded snprintf in DynvcTrackerHandleCreateBack then walks past the allocated buffer, leaking adjacent heap contents into proxy log output or crashing the process; only the proxy's own backend connection path is exposed, so environments that connect solely to fully trusted RDP servers are not practically at risk. No public exploit has been identified at time of analysis, EPSS is not elevated, and the issue is not listed in CISA KEV; impact is bounded to memory disclosure and denial of service rather than code execution.

Malicious or compromised RD Gateway servers can crash FreeRDP clients running versions 3.0.0 up to but not including 3.32.0 by returning consent or service messages whose MsgBytes field is set far larger than the NDR string ActualCount, causing the UTF-8 conversion routine to read past the end of the stream allocation. The defect is a client-side heap out-of-bounds read (CWE-125) with a CVSS score of 5.3; the CVSS 4.0 vector indicates PR:N but UI:P, meaning the attacker needs no credentials on the client yet the victim must actively initiate an RDP session through the attacker-controlled gateway. The only confirmed outcome is a client-process crash - there is no code execution or data exposure - and at time of analysis there is no public exploit identified and the issue is not in CISA KEV; a vendor fix is available in FreeRDP 3.32.0.

FreeRDP clients on Unix built with PKCS#11 support (3.0.0-beta1 up to but not including 3.32.0) can be crashed by a stack-based buffer overflow in WinPR's NCrypt PKCS#11 provider when an attacker-supplied PKCS#11 token or module presents a key and matching certificate whose CKA_ID is between 55 and 64 bytes long and the victim triggers key enumeration, typically during smart-card authentication. The helper wprintKeyName() hex-encodes the CKA_ID into a fixed 128-byte stack buffer, so a 64-byte ID produces 128 characters with no room for the terminating NUL, and the write is not bounded by the destination size. Impact is confined to a client-side crash with no demonstrated code execution or data disclosure, and exploitability is sharply narrowed by three prerequisites: the non-default PKCS#11 build configuration, physical delivery of a malicious token/module to the victim, and user interaction to start enumeration; no public exploit code or active exploitation has been identified at time of analysis, and this is a low-priority issue despite the CWE-121 classification.

Heap-based out-of-bounds write in FreeRDP 3.18.0 through 3.31.x allows a malicious or compromised ARM routing service to crash the RDP client by returning a long redirectedUsername that overflows the CLI credential prompt buffer once the user types input. The vulnerability requires attacker control of the ARM/gateway redirection service in the connection path (AT:P) and passive user interaction (UI:P), and is unauthenticated (PR:N) per the CVSS 4.0 vector; successful exploitation results in a client-side denial of service with no confidentiality or integrity impact. No public exploit code or confirmed active exploitation has been identified at time of analysis.

Xen 4.22 hosts running the C implementation of xenstored (as daemon or stubdom) can have the entire Xenstore service crashed or returned with corrupted data when an unprivileged guest holds a Xenstore watch with a specified watch depth and the host performs a Xenstore live update, producing a host-level denial of service for further domain administration actions. The root problem is that the watch state record is serialized two bytes short, so the newly booted xenstored misreads the record and faults. Exploitation is heavily gated on operator behavior - it requires Xen 4.22 plus C xenstored, a guest-registered watch with a stated depth, and an actual live update occurring while that watch is live - consistent with the assessed vector CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:C/C:N/I:L/A:H; no public exploit code exists and the issue is not in CISA KEV, with no EPSS or POC signals present in the available data. Only Xen 4.22 with C xenstored is affected; Xen 4.21 and earlier and installations using oxenstored are not vulnerable.

Heap-based buffer overflow in NVIDIA Warp crashes the host process when a user locally opens, imports, or runs a crafted simulation or data payload, producing a denial-of-service condition. The flaw is only reachable with local processing of attacker-controlled input plus user interaction (CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:L/A:L, 4.4 Medium), so it cannot be triggered across the network and no path to code execution or data disclosure has been demonstrated - impact is confined to availability. No public exploit code identified at time of analysis, and the issue is not listed in CISA KEV, so there is no confirmed active exploitation.

An integer overflow (CWE-190) in Firefox's Layout: Text and Fonts component can be triggered when a user renders attacker-controlled web content containing malformed text or font data, potentially causing memory-safety corruption in the browser process. The vulnerability affects Firefox builds prior to version 158 and Firefox ESR builds prior to 153.5, is reachable without authentication (CVSS:3.1/AV:N/AC:L/PR:N/UI:R), but requires the victim to load the crafted page, and its assessed impact is bounded to low confidentiality and integrity effects with no availability impact (score 5.4). Exploitation is not confirmed in the field: there is no CISA KEV listing and no public exploit identified at time of analysis, and the EPSS score of 0.15% (4th percentile) indicates a low predicted likelihood of exploitation activity.

Incorrect boundary conditions in Mozilla Firefox's Layout component prior to version 158 allow remote, unauthenticated attackers to cause low-severity confidentiality and integrity impact by enticing a user to render attacker-controlled web content; the assessed CVSS vector (AV:N/AC:L/PR:N/UI:R/S:U/C:L/I:L/A:N) confirms that user interaction is mandatory and that availability is not affected. No public exploit code has been identified and there is no CISA KEV entry at time of analysis, while the EPSS score is low at 0.15% (4th percentile), consistent with a moderate, not urgent, priority. The flaw is a memory-buffer boundary error (CWE-119) fixed in Firefox 158.

Crashing of the clevis-pkcs11-afunix-socket-unlock daemon in the clevis PKCS#11 pin allows a local attacker to cause a boot-time denial of service on hosts that rely on clevis for LUKS disk decryption. When a client connects to the daemon's AF_UNIX data socket without binding a client address, an integer underflow in the getpeername() result handling corrupts memory (CWE-787) and terminates the daemon, so LUKS unlock never completes and the system cannot finish booting. The issue is availability-only (CVSS 3.1 base 4.7, AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:H), affects clevis on Red Hat Enterprise Linux 7, 8, 9 and 10 as well as Red Hat OpenShift Container Platform 4, and no public exploit has been identified at time of analysis; the vulnerable configuration is a non-default deployment choice and not present in a stock RHEL install.

A stack-based out-of-bounds write in GNOME Remote Desktop's VNC TLS read path, as shipped in Red Hat Enterprise Linux 8, 9, and 10, can crash the service and cause denial of service. An unauthenticated remote attacker (per the CVSS vector PR:N) who can reach an enabled VNC TLS listener and drive the connection into a peek/read size mismatch can corrupt memory on the stack; the vector rates this CVSS 5.9 with availability-only impact (C:N/I:N/A:H) and high attack complexity (AC:H), reflecting that the mismatch is non-trivial to trigger reliably. The vulnerable configuration is not default: remote desktop sharing must be explicitly enabled with the VNC backend and a reachable TLS listener, which is uncommon on headless or server RHEL installs; no public exploit code and no confirmed active exploitation were identified at time of analysis, and this should be handled as a routine patch rather than an emergency.

Out-of-bounds memory reads in Siemens Desigo PXC building-automation controllers (all versions >= V2.3) and Capital Embedded AR Classic (431-422 all versions; R20-11 versions before V2303) allow an unauthenticated, remote attacker who can reach the device's packet-processing/communication service to read memory outside the intended packet boundaries by sending a malformed packet. Per the assessed CVSS 3.1 vector AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N (score 5.3), the impact is limited to low confidentiality loss - no integrity or availability impact, and no code execution - because the flaw is a pure out-of-bounds read (CWE-125). There is no public exploit identified at time of analysis and no CISA KEV listing; the practical risk is governed by network exposure of these typically segmented OT/building-management interfaces.

An out-of-bounds read in Tiiny-AI PowerInfer's CUDA backend lets a local, authenticated user crash or partially disclose memory by supplying a crafted model or matrix input to the FFN GPU-split loader path (create_striped_mat_to_gpu in smallthinker/ggml/src/ggml-cuda/ggml-cuda.cu). Exploitation requires an existing low-privileged local account and a CUDA-capable deployment that actually exercises the GPU-split FFN loading path; there is no remote vector and impact is confined to the local process (CWE-125, CVSS 4.0 base 1.9, assessed 3.1 AV:L/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:L). Publicly available exploit code exists (referenced via the project's GitHub issue tracker), but there is no confirmed active exploitation (no CISA KEV listing), and the overall risk is low. PowerInfer uses a rolling release model, so no affected or fixed version numbers are disclosed, and the proposed fix pull request awaits acceptance.

An out-of-bounds read in the GGUF model parser of Tiiny-AI PowerInfer (gguf_init_from_file in ggml.c) lets a local attacker who can supply a crafted GGUF file trigger the parser to read past the end of a buffer via a manipulated data_offset value, resulting in memory disclosure or a crash rather than code execution. Only local users with low privileges (CVSS PR:L, AV:L, no user interaction) are able to reach the vulnerable code path, and the attacker must control or substitute the model file that PowerInfer loads; the issue is rated only 1.9 under CVSS 4.0, and our independent assessment aligns with a Low/C:L/I:N/A:L profile on CVSS 3.1. Publicly available exploit code exists (the reported issue includes a proof-of-concept), but there is no confirmed active exploitation (not listed in CISA KEV), and no accepted upstream patch - a pull request addressing the flaw is still awaiting acceptance.

Loading a malformed or tampered sherpa-onnx offline transducer model pack (versions up to and including 1.13.8) can trigger a buffer overflow in OfflineTransducerGreedySearchDecoder::Decode, because the decoder trusts an unvalidated vocab_size value taken from model metadata to index decoder buffers. The issue is exploited by supplying a model pack whose metadata vocab_size is inconsistent with the joiner's actual output dimension; the attack requires the victim to load that attacker-supplied model, so the CVSS 4.0 vector carries UI:P and our independent assessment rates it CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:L/I:L/A:L with only partial confidentiality/integrity/availability impact. Publicly available exploit code exists via the upstream GitHub issue, but there is no evidence of confirmed active exploitation (not in CISA KEV), and the vendor notes that the bug is not reachable when only official model packs with metadata consistent with the joiner are used, making real-world risk LOW despite the 'remote buffer overflow' framing and the low EPSS-style base score of 2.1.

Buffer Overflow vulnerability in zserge jsmn jsmn HEAD 25647e6 allows a remote attacker to cause a denial of service via the jsmn.h, jsmn_parse(), example/jsondump.c, dump() components

Memory corruption in the Zephyr RTOS Bluetooth host ATT transmit path can be triggered by a peer in Bluetooth radio range whenever an application issues an ATT Signed Write Command over an unencrypted link. In chan_send(), the BT_ATT_OP_SIGNED_WRITE_CMD branch unrefers a net_buf it does not own after bt_smp_sign() fails (no local CSRK exists for the peer - a state the remote peer controls by pairing without requesting the signing key, pairing with No Bonding, staying unpaired, or keeping the link unencrypted); the callers honor the opposite ownership contract, so the already-freed buffer is re-queued, dereferenced and freed a second time, corrupting the ATT TX FIFO, deferred-destroy and pool free lists, while a related retry path can append a second 12-octet signature into a pooled buffer sized for one. Exploitation requires the deprecated, non-default CONFIG_BT_SIGNING build option plus a victim application that actually performs signed writes, giving CVSS 5.9 (AV:A/AC:H/PR:N, no confidentiality impact, high availability impact, low integrity impact); a vendor fix commit and GHSA-p3r4-qw98-j59m are available, and no public exploit was identified at time of analysis.

Unauthenticated flooding of crafted DHCPv6 RENEW messages can repeatedly crash odhcpd, the DHCPv6/DHCPv4/router-advertisement server shipped and enabled by default on OpenWrt, denying IPv6 address configuration to clients on the local link. An attacker who is network-adjacent (AV:A, PR:N, UI:N) must first complete two legitimate SOLICIT/REQUEST exchanges to obtain bindings, then send a single RENEW containing multiple IA_NA options that triggers an unsigned size_t underflow and lets attacker-controlled data be serialized past the 512-byte stack buffer pdbuf; the stack canary terminates the daemon and procd respawns it, so sustained packets cause recurring DHCPv6 outages. Impact is availability-only (C:N/I:N/A:H) and the observed configuration does not establish code execution; no public exploit identified at time of analysis, and the issue is fixed in OpenWrt 25.12.5.

An out-of-bounds heap write in Moonlight Qt 6.1.0 and earlier allows an attacker-controlled GameStream host to corrupt memory in the Moonlight client and reliably crash it during an active streaming session. Exploitation requires the victim to voluntarily initiate a session to the malicious host and requires Moonlight to select a decoder path that does not advertise CAPABILITY_REFERENCE_FRAME_INVALIDATION_AVC, which sends the server's H.264 SPS data through the vulnerable fixup path in app/streaming/video/ffmpeg.cpp, where a seq_parameter_set_id of 32 or greater is used unvalidated to index the 32-entry sps_table. The assessed impact is availability-only client crash and heap corruption (CVSS 6.5, AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H); despite metadata tags referencing RCE, code execution has not been established, and no public exploit code has been identified at time of analysis. This is a medium-severity, client-side denial-of-service issue rather than a high-priority remote compromise.

Loading an attacker-supplied read-only dictionary (.rws) file in GNU Aspell through 0.60.8.2 can corrupt heap memory, yielding reliable denial of service and potentially code execution. The root cause is a heap-based buffer overflow write (CWE-122) triggered by a signed-integer overflow in String::reserve(), which lets a crafted 32-bit length field in the .rws header (for example lang_name_size = 0x7FFFFFFF, paired with the valid endian marker 12345678) bypass the capacity check so attacker-controlled data is written past an undersized allocation. Exploitation is local in nature - there is no network-service trigger - and requires the victim to actively load the malicious dictionary, such as by running 'aspell dump master <file>'; no public exploit code was identified and there is no confirmed active exploitation at time of analysis, so this is assessed as a moderate-priority issue despite the RCE tag.

Denial of service in IBM Db2 11.5.0 through 11.5.9 and 12.1.0 through 12.1.5 can be triggered by an authenticated, low-privileged user (PR:L) submitting a specially crafted query that references a Materialized Query Table, causing a stack-based buffer overflow (CWE-121) that crashes the database engine. IBM has released a patch; no public exploit identified at time of analysis, and impact is limited to availability with no data disclosure or modification. Note that IBM's CVSS vector indicates local access (AV:L), while the query-execution nature suggests a remote authenticated client may be able to trigger the flaw, pending confirmation with the vendor.

An out-of-bounds write in the SPI host command interface of Semtech's LR11xx transceiver firmware lets an attacker who already has hands-on hardware access and logical control of the SPI bus send a crafted host command with an oversized length field, causing the firmware to write up to 254 arbitrary bytes beyond an internal non-persistent memory buffer. The affected product lines are the LR1110 transceiver and modem-E, the LR1120 transceiver, and the LR1121 transceiver and modem-E; the vulnerability is rated 5.2 (CVSS 4.0) and assessed at CVSS 3.1 AV:P/AC:L/PR:N/UI:N/S:U/C:N/I:H/A:H, meaning integrity and availability of the transceiver can be corrupted but no confidentiality loss occurs. There is no network or remote attack vector - exploitation requires physical access to the device plus SPI bus access - and no public exploit identified at time of analysis, with no confirmed active exploitation (CISA KEV) reported. Because the corrupted buffer is non-persistent, effects may be cleared by a device reset, keeping real-world risk low to moderate.

Heap buffer over-read in ImageMagick's -fx expression processor affects versions 7.1.0-21 through 7.1.2-31, allowing a local, low-privileged attacker who can get a crafted -fx expression evaluated to read past the end of an allocated heap buffer and disclose limited memory contents or trigger a crash. The issue is rated CVSS 4.0 4.8 (AV:L/AC:L/PR:L/UI:N, confined to low confidentiality and availability impact), and there is no public exploit identified at time of analysis. Practical risk is highest where a service passes untrusted input into ImageMagick -fx expressions and is comparatively small for typical operator-driven command-line use, matching the independent assessment that this is a low-to-moderate priority issue rather than a high-severity emergency.

ImageMagick 7.1.0-20 through 7.1.2-31 can crash or corrupt memory through a use-after-free in the MagickCore -fx expression evaluator, which is reached when the internal RPN element table must grow and the underlying memory allocation fails. Exploitation is constrained: it requires a local, low-privileged caller who can supply crafted -fx expressions to a process that is simultaneously under genuine memory pressure (the AT:P precondition in the CVSS 4.0 vector, base score 2.0), and the independent assessment caps impact at low availability loss with no confidentiality or integrity effect. The condition is not attacker-deterministic under ordinary conditions, no public exploit code has been identified at time of analysis, and this is not confirmed actively exploited (no CISA KEV listing), making it a genuinely low-priority issue whose signals are consistent rather than conflicting.

An uninitialized-memory information disclosure in ImageMagick's JPEG decoder affects all 6.x releases before 6.9.13-57 and 7.x releases before 7.1.2-32, allowing a remote, unauthenticated attacker to leak process memory by submitting a crafted JPEG for processing. The flaw is triggered when decoding a JPEG whose component count is fewer than four (for example single-component grayscale or 3-component YCbCr images), causing the decoder to read uninitialized fields from the JPEG component array; the leaked bytes are surfaced through the image's sampling-factor property, so exposure depends on whether the hosting application echoes image metadata back to the attacker. Exploitability is limited and disclosure reachability is conditional; there is no confirmed active exploitation (not in CISA KEV) and no public exploit code identified at time of analysis, though the vendor advisories and fixing commits are public. Vendor-released patches exist in 6.9.13-57 and 7.1.2-32, making this a moderate-priority remediation item rather than an emergency.

Malformed XMP metadata profiles in ImageMagick builds from 7.1.1-7 up to (but not including) 7.1.2-32 cause a heap buffer over-read that lets an attacker read past the end of an allocated buffer and typically crash the process. The flaw sits in the XMP profile synchronization code and is reachable by unauthenticated remote attackers (CVSS:4.0 PR:N/UI:N) whenever a service parses an untrusted image - upload handlers, thumbnailers, format converters, PDF/report generators, or any pipeline using an affected library build. Impact is limited to availability (VC:N/VI:N/VA:L, independently assessed as CVSS:3.1 AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L), so the realistic outcome is a denial-of-service crash rather than code execution or confirmed data disclosure; no public exploit code or CISA KEV entry has been identified at time of analysis, and a vendor patch is available.

An invalid memory free in ImageMagick's fx expression evaluator (MagickCore/fx.c) lets a local attacker with low privileges crash the process by forcing initialization of an FX runtime entry to fail, producing a denial-of-service condition in applications that pass attacker-controlled input to the -fx operator. Only the error-cleanup path is defective - the code frees fxrts[t] instead of fxrts[t2] - so the bug does not trigger during normal FX usage or default processing of benign images, and there is no confidentiality or integrity impact. No public exploit code has been identified at time of analysis and the issue is not referenced by CISA KEV; the independent assessment rates it low (CVSS 3.1 AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:L), and vendor fixes are available.

Client-side heap buffer overflow in FreeRDP 3.10.0 through 3.32.2 allows a malicious RDP peer to crash the connecting client by sending an oversized CF_DIB clipboard image whose conversion to PNG triggers an out-of-bounds memcpy once an internal realloc() call fails. The impact is availability-only (no code execution or data disclosure), which is why the issue is rated MEDIUM (CVSS 6.0) despite the memory-corruption class; the client is reached unauthenticated over the network, but the victim must actually interact with the clipboard transfer and the overflow only fires when a memory-allocation failure occurs, so the bug is not deterministic. There is no public exploit identified at time of analysis and no CISA KEV listing, and the affected range is limited to FreeRDP 3.10.0-3.32.2 with the client clipboard path as the attack surface.

Heap buffer overflow in FreeRDP's RDP license negotiation path lets a malicious or man-in-the-middle RDP server corrupt client heap memory when a 32-bit FreeRDP client (3.0.0 through the 3.31.x line, fixed in 3.32.0) connects to it and receives a crafted SERVER_LICENSE_REQUEST carrying an oversized scopeCount that defeats the length check. A successful attack can crash or destabilize the client (denial of service) and could potentially lead to code execution in the client's context, but exploitation requires the victim to initiate an outbound RDP session to an attacker-controlled endpoint and the integer truncation only manifests on 32-bit builds, not 64-bit ones. No public exploit identified at time of analysis; the vendor fix is shipped in FreeRDP 3.32.0, and the assessed risk vector is CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:N/I:H/A:H.