Red Hat
Monthly
The full-scope-disabled client-policy executor in Red Hat Build of Keycloak silently skips enforcement when the fullScopeAllowed attribute is absent from a client registration or update request, allowing a delegated user to create OAuth/OIDC clients with unrestricted scope access. Affected platforms span Red Hat Build of Keycloak, Red Hat Single Sign-On 7, Red Hat Data Grid 8, and JBoss EAP Expansion Pack. An attacker holding delegated registration privileges (CVSS PR:L) can exploit this to issue tokens containing unauthorized role mappings, effectively escalating their effective permissions within any application trusting those tokens. No public exploit code or CISA KEV entry has been identified at time of analysis.
pgAdmin 4 SERVER mode (versions 9.3 through 9.16) allows authenticated users to fully bypass administrator-configured per-tool role-based access controls by exploiting incomplete permission enforcement across backend routes and Socket.IO handlers. The Flask-Security `permissions_required` decorator was applied only to a single front-door route per tool, while all secondary routes and Socket.IO event handlers checked only authentication, not tool permissions - meaning a user who received a 403 on the initial gated endpoint could still drive the complete tool workflow through its other endpoints. Affected tools include Query Tool, Grant Wizard, Schema Diff, ERD, PSQL (interactive terminal), Debugger, Backup, Restore, Maintenance, and Import/Export. No dedicated public exploit has been identified, though the fix commits and GitHub issue are publicly visible; exploitation requires a valid authenticated pgAdmin session.
Denial of service in Apache HttpComponents Client (versions 5.0-alpha1 through 5.6.2) arises when the classic/blocking I/O HttpClient fails to release the underlying connection back to the pool after receiving a response carrying an invalid or unsupported Content-Encoding header. An attacker who controls or can influence the server a client talks to can repeatedly return such headers to leak connections until the pool is exhausted, halting further outbound HTTP requests from the affected Java application. The async I/O HttpClient is explicitly unaffected; there is no public exploit identified at time of analysis and no CVSS or EPSS data was supplied.
SAML HTTP-Redirect binding in Keycloak allows network-based attackers to inject malicious parameters into authentication requests targeting clients configured with wildcard redirect URLs. When a victim user completes authentication, Keycloak appends its legitimate SAML response to the attacker-controlled URL parameters, which can cause susceptible service providers to process the attacker's data instead of the real login assertion - potentially resulting in the victim being logged into the wrong account. No public exploit is identified at time of analysis, and the CVSS score of 3.4 (Low) reflects the high attack complexity and limited integrity-only impact.
Insecure file permissions in ansible-collection-redhat-leapp expose PostgreSQL backup archives to local non-root users on managed RHEL 9 and RHEL 10 nodes. When the `leapp_old_postgresql_data` remediation option is invoked with elevated privileges, the resulting backup archive is written with world-readable or group-readable permissions (CWE-732), allowing any local non-root user on that managed node to extract the full contents of the archived PostgreSQL data. No public exploit has been identified at time of analysis, and exploitation requires local authenticated access to the managed node.
The ansible-collection-redhat-leapp package exposes Ansible controller-side sensitive files to attacker-controlled managed nodes through a manipulated Leapp upgrade report. An attacker who has already obtained privileged write access to a managed node's Leapp report can craft malicious report content that, when an operator executes a specific remediation task, causes the Ansible controller to read and copy its own local files - including private keys and credentials - to the attacker-controlled managed node. No public exploit code has been identified at time of analysis, and this vulnerability is not listed in the CISA KEV catalog; however, the CVSS 3.1 score of 6.2 reflects real-world risk limited by the high privilege prerequisite and required operator interaction.
Uncontrolled decompression in aiohttp's WebSocket client prior to version 3.14.2 allows a malicious server to exhaust CPU and memory on connecting clients by sending RSV1-flagged compressed frames without having negotiated the permessage-deflate extension. The root cause is a missing constructor argument in `client.py:_ws_connect` that caused `WebSocketReader` to default to `compress=True`, violating RFC 6455 §5.2 which mandates failing the connection on undefined reserved bits. This is client-side exposure only - aiohttp server deployments are not affected. No active exploitation has been identified and no public exploit code exists at time of analysis.
Denial of service in Samba's Kerberos KDC password change (kpasswd) service allows authenticated remote attackers to crash the KDC process by submitting specially crafted, malformed ASN.1-encoded password change requests. The server miscalculates the structure size of the incoming request and reads up to six bytes beyond the end of the allocated buffer; when this read strikes unmapped memory, the KDC process terminates rather than simply failing gracefully. No public exploit has been identified and no CISA KEV listing exists, but the disruption of a KDC in an Active Directory or Samba domain environment carries significant operational impact since authentication for the entire domain depends on KDC availability.
Samba's internal DNS server inserts unauthenticated TKEY registration requests into the TKEY name cache before rejecting them, allowing a remote unauthenticated attacker to exhaust the cache by flooding it with arbitrary TKEY names. Eviction of legitimate cache entries blocks TSIG-authenticated DNS query processing, producing a targeted denial of service against Active Directory environments relying on Samba's built-in DNS backend. No public exploit code or CISA KEV listing has been identified at time of analysis; a vendor patch is available from both Samba upstream and Red Hat.
Credential exfiltration in koku-metrics-operator allows a privileged user with edit access to the CostManagementMetricsConfig custom resource to redirect the operator's OAuth token requests to an attacker-controlled endpoint, exposing the tenant's Red Hat SSO client_id and client_secret. The operator's service-account authentication mode unconditionally forwards these credentials to whatever OAuth token URL is configured in the CR, acting as a confused deputy on behalf of the CR editor. No public exploit code has been identified and the vulnerability is not listed in CISA KEV, but the CVSS Scope:Changed metric reflects that the operator process holds credentials inaccessible to the CR editor directly, amplifying impact beyond the initial access tier.
Bearer-token disclosure in the Red Hat koku-metrics-operator lets a user with edit rights over the CostManagementMetricsConfig custom resource redirect metric uploads to an attacker-controlled URL, causing the operator to attach the cluster-global Red Hat Cloud pull-secret token to that outbound request and leak it. Because token authentication is the default (authentication.type=token) and the pull secret is shared cluster-wide, a single privileged tenant can harvest credentials that grant access to the organization's Red Hat Cloud/console.redhat.com services. There is no public exploit identified at time of analysis and it is not listed in CISA KEV; the issue was reported directly by Red Hat's security team.
Stack exhaustion in PHP's phar archive handler allows an attacker to crash the PHP process by supplying a crafted archive containing circular symbolic links, triggering unbounded recursion that depletes the native C stack. Affected versions span PHP 8.2.x before 8.2.33, 8.3.x before 8.3.33, 8.4.x before 8.4.24, and 8.5.x before 8.5.9 across all platforms. Exploitation requires a PHP process to actively parse the malicious archive (UI:A/AV:L); no public exploit has been identified and exploitation likelihood is rated unlikely (E:U in the provided CVSS 4.0 vector), making this a moderate-priority patch for most deployments - with heightened urgency for applications that process user-supplied phar archives.
Uninitialized memory use in Skia, Chrome's 2D graphics rendering engine, exposes process memory contents to remote attackers on Windows via crafted HTML pages in Google Chrome versions prior to 151.0.7922.72. The flaw is strictly an information disclosure issue (CWE-457), with no integrity or availability impact, affecting only the Windows platform per the vulnerability description. No public exploit code has been identified, EPSS places exploitation probability at 0.19% (9th percentile), and SSVC classifies exploitation as none with partial technical impact - collectively suggesting limited real-world risk despite the CVSS C:H rating.
Side-channel GPU information leakage in Google Chrome on Android prior to 151.0.7922.72 enables remote attackers to extract cross-origin data by serving a crafted HTML page. The vulnerability exploits the GPU rendering pipeline to observe timing or state artifacts that escape the Same-Origin Policy boundary, allowing partial disclosure of data from other origins loaded in the browser. No active exploitation has been confirmed (not in CISA KEV), and the EPSS score of 0.17% (7th percentile) reflects low observed exploitation probability; a vendor-released patch is available.
Same-origin policy bypass in Google Chrome's DevTools component prior to version 151.0.7922.72 allows an unauthenticated remote attacker to perform unauthorized cross-origin write operations against a victim's browser context. Exploitation requires the victim to visit a specially crafted HTML page (UI:R), making it a socially-engineered threat rather than a purely opportunistic one. EPSS is low at 0.22% (13th percentile) and no CISA KEV listing or public exploit code has been identified, consistent with the Chromium team's own 'Medium' severity rating; however, Chrome's ubiquitous deployment makes even medium-severity findings broadly relevant.
Cross-origin data leakage in Google Chrome's WebProtect component (prior to 151.0.7922.72) enables a remote attacker who has already compromised the renderer process to exfiltrate cross-origin data by loading a crafted HTML page. The flaw stems from an inappropriate implementation of origin validation (CWE-346), allowing WebProtect to fail its primary function of enforcing same-origin boundaries within Chrome's sandboxed renderer. No active exploitation has been confirmed by CISA KEV, no public POC has been identified, and EPSS at 0.21% (11th percentile) indicates low near-term exploitation probability - though this vulnerability's relevance as a component in chained browser exploit chains elevates its strategic significance beyond the raw score.
Cross-origin data leakage in Google Chrome's GPU implementation prior to version 151.0.7922.72 allows a remote attacker who has already compromised the renderer process to read cross-origin rendered content via a crafted HTML page. The flaw stems from an origin validation error (CWE-346) in the GPU compositor layer, enabling sandbox-escaped renderer processes to access pixel or texture data belonging to other security origins. EPSS at 0.21% (11th percentile) and no CISA KEV listing indicate low current exploitation probability; no public exploit code has been identified at time of analysis.
Cross-origin data leakage in Google Chrome's NoStatePrefetch feature exposes confidential resources across origin boundaries when a victim visits an attacker-crafted HTML page, affecting all Chrome versions prior to 151.0.7922.72. The flaw exploits observable side-channel signals produced during background prefetch operations to infer cross-origin state, violating Same-Origin Policy protections. No public exploit or active exploitation has been identified; EPSS sits at 0.21% (11th percentile), indicating low near-term exploitation likelihood despite the product's ubiquitous deployment footprint.
Cross-origin data leakage in Google Chrome's Skia graphics engine affects all Chrome versions prior to 151.0.7922.72, enabling remote attackers to read sensitive data across browser security boundaries when a victim visits a specially crafted HTML page. The root cause is uninitialized memory use (CWE-457) within the Skia 2D rendering library, which may expose residual memory contents - potentially including fragments from other origins - violating the Same-Origin Policy at the memory level. No public exploit code has been identified and EPSS sits at 0.23% (14th percentile), indicating low exploitation probability at time of analysis despite the vulnerability being remotely triggerable without authentication.
Navigation restriction bypass in Google Chrome's Presentation component (all versions prior to 151.0.7922.72) enables a remote attacker to circumvent enforced browser navigation policies via a specially crafted HTML page. The flaw carries a high integrity impact (I:H) with no privilege requirement on the attacker's side, though victim interaction is mandatory. EPSS sits at 0.22% (13th percentile) and SSVC classifies exploitation as 'none' with a non-automatable profile, placing this in the patch-but-not-emergency tier for most organizations. No public exploit code identified at time of analysis.
UI spoofing in Google Chrome's Extensions component enables an attacker who persuades a victim to install a malicious extension to misrepresent browser security indicators, affecting all Chrome desktop versions prior to 151.0.7922.72. The flaw (CWE-451) stems from incorrect rendering of security UI elements within the Extensions subsystem, granting a crafted extension the ability to deceive users about their browser's security state or extension permissions. No public exploit identified at time of analysis, and EPSS at 0.13% (3rd percentile) reflects very low exploitation probability; this CVE is not listed in CISA KEV.
Same-origin policy bypass in Google Chrome's Blink rendering engine (versions prior to 151.0.7922.72) allows a remote attacker to violate cross-origin isolation boundaries by luring a user to visit a specially crafted HTML page. The CVSS vector assigns High integrity impact with no privilege requirement, meaning successful exploitation could enable an attacker to inject or manipulate cross-origin content within the victim's browser session. No public exploit has been identified at time of analysis and EPSS sits at 0.22% (13th percentile), with SSVC confirming no observed exploitation - placing this in the patch-on-schedule rather than emergency-response tier.
Cross-origin data leakage in Google Chrome's Autofill subsystem (all versions prior to 151.0.7922.72) enables a remote attacker to read sensitive form data across origins by serving a crafted HTML page to the victim. The root cause is an origin validation error (CWE-346) where the Autofill implementation fails to enforce proper origin boundaries, violating the browser same-origin policy for autofill-populated fields. No public exploit code has been identified and the EPSS score of 0.21% places this in the 11th percentile, indicating very low current exploitation probability; a vendor patch is available.
Cross-origin data leakage in Google Chrome's Payments subsystem prior to version 151.0.7922.72 allows a remote attacker to exfiltrate limited cross-origin information by luring a victim to a crafted HTML page. The root cause is insufficient policy enforcement (CWE-346) within the Payments component, meaning Chrome fails to properly validate origin boundaries when handling payment-related operations. No public exploit code has been identified and EPSS is very low (0.21%), but the ubiquitous deployment of Chrome elevates aggregate exposure.
Cross-origin data leakage in Google Chrome's ANGLE graphics component (versions prior to 151.0.7922.72) exposes residual memory contents to remote attackers via a crafted HTML page. ANGLE - Chrome's graphics abstraction layer - fails to initialize memory before use (CWE-457), allowing pixel or buffer data from one origin to be read by another. No public exploit identified at time of analysis; EPSS probability is 0.23% (14th percentile), indicating low current exploitation activity despite Chrome's massive deployment footprint.
Side-channel information leakage in Google Chrome's Autofill feature (versions prior to 151.0.7922.72) enables cross-origin data exfiltration by exploiting observable behavioral differences during Autofill interactions. An unauthenticated remote attacker who delivers a crafted HTML page and persuades the victim to perform specific UI gestures can infer data belonging to other origins, bypassing Same-Origin Policy protections. EPSS is very low at 0.21% (11th percentile) and no active exploitation is confirmed via CISA KEV, making this a real but modest real-world risk - primarily meaningful where sensitive Autofill data (saved addresses, payment details) is stored in the target browser profile.
Cross-origin data leakage in Google Chrome's Actor component (all versions prior to 151.0.7922.72) allows a remote attacker who has already compromised the renderer process to exfiltrate sensitive cross-origin data via a crafted HTML page. This is a secondary-stage information disclosure primitive - it requires both user interaction and a pre-compromised renderer, making it a chained exploit component rather than a standalone entry point. No public exploit or active exploitation has been identified; EPSS sits at 0.29% (21st percentile), consistent with low real-world exploitation likelihood at this time.
Side-channel information leakage in Google Chrome's Media component (versions prior to 151.0.7922.72) enables remote attackers to exfiltrate cross-origin data by luring a victim to a crafted HTML page that exploits observable timing or state signals in media processing. This bypasses the browser's Same-Origin Policy protections, allowing an attacker's origin to infer sensitive data loaded from a different origin. No public exploit has been identified at time of analysis, and the EPSS score of 0.29% (21st percentile) reflects low current exploitation probability; a vendor-released patch is available.
Cross-origin data leakage in Google Chrome's FileSystem component prior to version 151.0.7922.72 enables a remote attacker to read data across origin boundaries by luring a victim to a crafted HTML page. The root cause is insufficient policy enforcement (CWE-346 Origin Validation Error) in Chrome's FileSystem API, which fails to enforce same-origin policy checks correctly, exposing data that should be sandboxed to its originating origin. No public exploit code has been identified at time of analysis, and the EPSS score of 0.29% (21st percentile) reflects low current exploitation likelihood; however, the vendor patch is available and should be applied promptly given Chrome's ubiquitous deployment footprint.
Uninitialized memory use in ANGLE, Chrome's graphics translation layer, exposes cross-origin data to remote attackers who can direct users to a crafted HTML page. Affected are all Google Chrome desktop releases prior to 151.0.7922.72 across supported platforms. No public exploit code or active exploitation has been identified; EPSS probability is low at 0.31% (24th percentile), and the impact is bounded to partial confidentiality loss with no integrity or availability consequences.
The BGR decoder in ImageMagick prior to 7.1.2-27 fails to check for end-of-file at all necessary code paths, allowing remote unauthenticated attackers to trigger a heap buffer over-read via a crafted image file. The CVSS vector (AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N) confirms network-exploitable, zero-prerequisite access with impact limited to partial memory disclosure. No active exploitation (CISA KEV) and no public POC have been identified at the time of analysis; a vendor-released patch is available in version 7.1.2-27.
CRLF injection in undici's HTTP/1.1 dispatcher enables HTTP header injection and potential request smuggling when applications pass duck-typed blob-like bodies whose type property is populated from untrusted input. Affected are the 6.x branch before 6.28.0, all 7.x releases from 7.0.0 through 7.28.x, and all 8.x releases from 8.0.0 through 8.8.x, when callers use the request, stream, pipeline, or dispatch APIs with hand-rolled blob-shaped objects rather than native Blob instances or the fetch API. An attacker who controls the type value can append arbitrary HTTP headers to the outgoing request, potentially smuggling a second request past an upstream proxy or server. No public exploit identified at time of analysis and KEV status is not confirmed in the provided data.
Response body/Content-Length desynchronization in undici's retry interceptor allows a malicious or faulty upstream server to induce HTTP response smuggling against downstream clients of proxy and gateway applications. Affected versions span undici before 6.28.0, the 7.x line before 7.29.0, and the 8.x line before 8.9.0. An upstream that deliberately returns a partial response with mismatched framing and closes the socket early can cause the retry interceptor to assemble a body of a different length than the attached Content-Length, leading downstream consumers to experience response desynchronization, connection hangs, or data corruption. No public exploit has been identified at time of analysis, and this is not listed in CISA KEV.
Server-Side Request Forgery in Pydantic AI's UI adapter layer (versions 1.65.0-1.105.0 and 2.0.0b1-2.0.0b5) allows unauthenticated network clients to redirect the application server into fetching arbitrary cloud-storage objects using the server's own credentials. UploadedFile references - carrying provider file IDs or cloud-storage URIs such as s3:// or gs:// - submitted via client message history bypass the scheme allowlist applied to ordinary file URLs and are resolved server-side using the application's IAM role, service account, or API key. No public exploit has been identified at time of analysis, and no CISA KEV listing exists, but the scope-changing confidentiality impact (CVSS S:C/C:H) warrants prompt patching for any deployment exposing the Vercel AI adapter or similar UI adapters to untrusted clients.
Tool call injection in Pydantic AI's UI adapters (AG-UI via Agent.to_ag_ui()/AGUIAdapter, and Vercel AI via VercelAIAdapter) enables remote unauthenticated clients to force execution of registered server-side tools with attacker-supplied arguments, bypassing the framework's message-sanitization defense. The flaw affects pydantic-ai and pydantic-ai-slim versions 1.88.0-1.107.0 and 2.0.0b1-2.4.x. No public exploit code or CISA KEV listing exists at time of analysis, but the attack is low-complexity and exploitable against default configurations of applications using either UI adapter.
Arbitrary file overwrite in Develar app-builder (all versions ≤ 4.2.0 on macOS) allows an unauthenticated remote attacker to escape the ZIP extraction directory by exploiting APFS Unicode case-folding combined with symlink following in the `zipx.Unzip` routine. A crafted archive containing a symlink entry `ss` pointing to an external victim file and a subsequent regular-file entry `ß` causes the builder to follow the symlink - because APFS folds `ß` to `ss` - overwriting the target with attacker-controlled content. SSVC rates the attack automatable with no confirmed exploitation; no public exploit code has been identified at time of analysis, though CERT/CC advisory VU#293714 and ENISA EUVD-2026-50423 confirm coordinated disclosure.
Cookie attribute injection in undici's setCookie function allows attacker-controlled domain values or unparsed array entries to silently append, strip, or override attributes such as SameSite, HttpOnly, and Secure. Affected are all undici releases before 6.28.0, the 7.x branch before 7.29.0, and the 8.x branch before 8.9.0. In multi-tenant or reverse-proxy deployments that scope session cookies to tenant-supplied domains, this flaw enables CSRF protection bypass by downgrading or overriding the SameSite attribute, with no public exploit identified at time of analysis.
Node.js Permission Model fails to enforce filesystem write restrictions for the trace events subsystem, allowing a local attacker with low privileges to write files to paths explicitly blocked by the operator via the --experimental-permission flag. Rated Low severity (CVSS 3.3), this flaw affects Node.js 22.x through 22.23.1, 24.x through 24.18.0, and 26.x through 26.5.0, and is one of ten CVEs addressed in the July 2026 Node.js security release. No public exploit code has been identified and the vulnerability is not listed in CISA KEV; the fix is available in Node.js 26.5.1 released 2026-07-29.
IP-based security controls in applications using @fastify/forwarded before 3.0.2 can be bypassed by injecting a horizontal tab character into X-Forwarded-For header entries. The parser strips only space characters during whitespace normalization, violating RFC 7230's definition of optional whitespace (which includes both SP and HTAB), causing the resolved IP string to carry a trailing tab. Any application performing exact-string security decisions on the resolved address - allowlists, blocklists, per-IP rate limit keys, or audit log correlation - will silently fail to match the expected value, allowing an attacker to evade those controls. No public exploit has been identified at time of analysis, and no KEV listing exists; however, exploitation requires no authentication or special configuration beyond controlling HTTP request headers.
Non-atomic JSON Patch application in cJSON through 1.7.19 allows an attacker who controls the patch document to silently corrupt the target JSON structure even when the API returns a failure status. The apply_patch() function in cJSON_Utils.c deletes existing target members before completing operation validation, meaning a malformed replace (missing value) or move (unresolvable destination) permanently destroys data from the in-memory document tree - defeating the RFC 6902 all-or-nothing contract that callers rely on to reject bad patches safely. No public exploit has been identified at time of analysis, but a detailed researcher disclosure and direct source code reference are publicly available, which may lower the barrier to exploitation; KEV status is not confirmed.
Path traversal in open-iscsi allows remote MITM attackers to create root-owned files at arbitrary filesystem locations and inject lines into records by embedding RFC 3722-prohibited characters in IQN target names returned during iSNS or firmware discovery. All open-iscsi versions through commit 668ca1df9c9a1e9bdd5c999ae1d67c9c8909237e are affected across the iSNS discovery, firmware discovery, and command-line target-name paths. No public exploit code or CISA KEV listing has been identified at time of analysis; however, the CVSS 4.0 score of 6.9 with network vector and no privileges required warrants attention in enterprise storage environments where iSCSI discovery is exposed.
Terminal control sequence injection in diff-so-fancy exposes developers to output manipulation, filename spoofing, terminal screen clearing, and clipboard hijacking when viewing diffs containing attacker-controlled content. The application filters only ANSI SGR sequences while passing OSC, CSI, carriage returns, and other VT100 control characters through unsanitized to the terminal, affecting all versions prior to commit 9c81294. No public exploit is identified at time of analysis, but the low attack complexity (CVSS 4.0 AV:L/AC:L) means the primary barrier is only the requirement for a developer to actively view the malicious diff.
Unauthenticated disabling of the ACME service in Dogtag PKI allows persistent denial of service via URL manipulation. By appending a trailing slash to admin-only enable/disable endpoints, an attacker can bypass the Tomcat authentication constraint while RESTEasy routes the request, enabling unauthorized toggling. No public exploit or active exploitation has been reported.
Resource exhaustion in Ninenines Cowboy's HTTP/1.1 handler allows unauthenticated remote attackers to cause denial of service by sending many header lines with the same name, bypassing the max_headers limit due to counting distinct names instead of total lines. Affects Cowboy versions 2.0.0-pre.4 to 2.17.x; fixed in 2.18.0. No public exploit or active exploitation has been reported at this time.
Safari and WebKit-based apps across Apple platforms crash when a user visits a specially crafted website, due to a state management flaw (CWE-400: Uncontrolled Resource Consumption) in versions prior to 26.6. Affected platforms include Safari on macOS Tahoe, iOS 26, iPadOS 26, and visionOS 26. No confidentiality or integrity impact is present - the consequence is an app-level denial-of-service. No public exploit code exists and CISA SSVC confirms no active exploitation, with EPSS at 0.16% (6th percentile), indicating very low exploitation likelihood at time of analysis.
UI spoofing in Apple Safari 26.5 and earlier allows attackers using cross-origin framing to misrepresent website identity. The vulnerability requires user interaction (visiting a malicious page) and could lead to phishing or deceptive actions. No active exploitation or public exploit code identified; patch is available.
Bypass of iframe sandboxing policy in Apple Safari and all Apple operating systems through version 26.5 allows malicious web content to violate sandbox restrictions. Affected users include anyone running Safari on iOS, iPadOS, macOS Tahoe, tvOS, visionOS, or watchOS prior to version 26.6. An attacker can cause integrity violations, such as injecting or modifying content within sandboxed iframes, but requires user interaction (clicking a link or visiting a malicious page) and no public exploit or active exploitation has been identified at this time.
An integer overflow in GIMP's file-sgi plugin, triggered by a crafted RLE-compressed SGI image, causes denial of service by crashing the plugin. The vulnerability affects GIMP versions with the file-sgi plugin and requires a user to open a malicious image file. No public exploit or active exploitation is known at this time; the issue is being tracked by GNOME and Red Hat.
Path traversal vulnerability in Erlang OTP's stdlib zip module allows attackers to write files outside the intended extraction directory by exploiting flawed validation of relative paths in zip archives. Affected are OTP versions 27.1 before 27.3.4.15, 28.5.0.4, and 29.0.4. Successful exploitation requires user interaction (extracting a crafted zip) and has limited impact on integrity and availability; no active exploitation has been reported.
Remote denial of service in llama.cpp allows unauthenticated attackers to exhaust server resources via crafted JSON schema inputs that trigger unbounded recursion in the JSON-schema-to-GBNF conversion routine. Affected versions include commit e15efe0 and prior. A fix exists in a pending pull request; no active exploitation reported.
Denial of service in SQLite 3.41 due to a use-after-free in the JSON extraction function. A local unauthenticated attacker can trigger a crash by supplying a malicious JSON input, causing the host application to terminate. No active exploitation has been confirmed in the wild, but public exploit proof-of-concept may exist.
Memory exhaustion in ImageMagick's magick CLI tool allows a local attacker to cause denial of service by repeatedly supplying malformed command-line arguments. Affects versions before 7.1.2-27. Vendor patch is available; no public exploit or active exploitation reported.
NULL pointer dereference in the Linux kernel tracing subsystem (`func_set_flag()`) allows a local attacker with tracefs write access to trigger a kernel panic and crash the system. The flaw arises when a process holds an open file descriptor to a function tracer option (e.g., `func_stack_trace`) while a concurrent tracer switch sets `tr->current_trace_flags` to NULL; a subsequent write to that option file then dereferences the NULL pointer before the function tracer guard executes. No active exploitation (KEV) or public exploit code has been identified; EPSS is 0.20% at the 10th percentile, consistent with a localized, technically specific DoS condition.
Kernel crash (availability loss) in the Linux gpio/eic-sprd Spreadtrum EIC GPIO driver affects systems running PREEMPT_RT kernels with Unisoc/Spreadtrum SoCs. The IRQ startup path - request_threaded_irq() → irq_startup() → sprd_eic_irq_unmask() - calls sprd_eic_update(), which acquires a regular spinlock_t via spin_lock_irqsave(). On PREEMPT_RT kernels, spinlock_t is a sleeping lock; sleeping in a non-sleepable IRQ-disabled context triggers a kernel BUG: 'sleeping function called from invalid context', crashing the system. No public exploit has been identified at time of analysis; patches are available across all maintained stable branches.
Kernel panic risk in the Linux gpio-sch driver exposes PREEMPT_RT systems with legacy Intel SCH GPIO hardware to a local denial of service. The SCH IRQ unmask callback acquires a regular spinlock inside a non-sleepable IRQ-setup context; under PREEMPT_RT, that spinlock is reimplemented as a sleeping lock, triggering a lockdep BUG and potential kernel panic. Patches are available across all active stable branches (5.15.212, 6.1.178, 6.6.145, 6.12.97, 6.18.39, 7.1.4); no public exploit exists and EPSS is 0.21% (percentile 12%).
NULL pointer dereference in the Linux kernel's hid_logitech_dj driver causes a kernel crash (DoS) when a crafted HID descriptor is processed during device probe. The flaw exists in logi_dj_probe(): when a Logitech Unifying receiver emulation presents report ID 0x20 with only padding output items, hid-core registers the report with rep->maxfield == 0. The error-path validation then dereferences rep->field[0]->report_count while printing a diagnostic message, triggering a KASAN-visible null-ptr-deref and kernel panic. No public exploit is identified at time of analysis, but the kernel commit descriptions include a full uhid-based reproduction method, lowering the barrier for local attackers.
Resource exhaustion via file reference leak in the Linux kernel's io_uring NOP opcode handler allows a local, low-privileged user to gradually exhaust kernel memory. The flaw occurs when a NOP submission carries the generic IOSQE_FIXED_FILE sqe flag (setting REQ_F_FIXED_FILE) but omits the NOP-specific IORING_NOP_FIXED_FILE flag in nop_flags: io_nop() acquires a real file reference via io_file_get_normal(), yet io_put_file() skips releasing it because REQ_F_FIXED_FILE is set, producing one leaked 176-byte kernel file object per operation. No public exploit has been identified at time of analysis, and the vulnerability is not listed in the CISA KEV catalog.
Availability impact in the Linux kernel's io_uring work queue (io-wq) subsystem allows a local, low-privileged user to cause an extended hang during io_uring ring closure. The root cause is a race condition in io_worker_handle_work(): the IO_WQ_BIT_EXIT check is snapshotted once before processing a linked-work chain, so if io_wq_exit_start() fires mid-chain, remaining linked items execute with a stale do_kill=false, skip IO_WQ_WORK_CANCEL, and continue running after exit has been initiated. No public exploit is identified at time of analysis; the issue was surfaced by syzbot using slow MSR-device reads with linked work chains and immediate ring close.
Kernel availability loss in the Linux mm/shrinker subsystem allows a local user with debugfs access to trigger a kernel BUG by reading the debugfs 'count' file of a memcg-aware shrinker such as zswap. The shrinker_debugfs_count_show() function incorrectly holds rcu_read_lock() while invoking ->count_objects() callbacks; when zswap's callback attempts to flush memcg stats via css_rstat_flush(), it sleeps inside an RCU read-side critical section - a fatal kernel invariant violation. No public exploit has been identified at time of analysis, and EPSS is very low at 0.21% (12th percentile).
Kernel panic via NULL pointer dereference in the Linux kernel's memory shrinker debugfs interface allows a local attacker to crash the system. The shrinker_debugfs_add() function unconditionally creates 'count' and 'scan' debugfs files for all registered shrinkers, but some shrinkers - notably the xen-backend shrinker - only implement one of the two required callbacks (count_objects or scan_objects). Writing to the exposed 'scan' debugfs file on such a shrinker dispatches through a NULL function pointer, triggering a kernel panic. No public exploit has been identified at time of analysis, and EPSS score of 0.21% reflects low exploitation probability. Patches are available across multiple stable kernel branches.
NULL pointer dereference in the Linux kernel's swap cgroup subsystem (`mm/swap_cgroup.c`) causes a kernel panic and full system crash on swapless hosts running kernel versions after commit `bea67dcc5eea`. The batched PTE zap path introduced by that commit calls `lookup_swap_cgroup_id()` without validating the swap entry against `swap_info[]`, so any process exit involving a PTE corrupted into a type-0 swap entry dereferences the uninitialized `ctrl->map` pointer and takes the host down. The crash pattern was confirmed in production on kernel 6.12.58 and independently reproduced by syzbot; no public exploit exists, but the kernel teardown path is deterministically vulnerable once the prerequisite conditions are present.
Softlockup in the Linux kernel's swap memory reclaim path allows a local user to trigger a kernel panic by generating heavy swap pressure on systems with large swap devices. The swap_reclaim_full_clusters() function, called from a kworker context, iterates through potentially thousands of full swap clusters without yielding the CPU via cond_resched(), starving the scheduler and eventually causing the kernel watchdog timer to fire. The vulnerability was confirmed in a real-world stress test on a 320-CPU arm64 machine with ~1TB RAM and an 8.6GB swap device, resulting in a kernel panic. No public exploit code exists and EPSS is very low (0.21%, 11th percentile).
Denial of service in the Linux kernel btmtksdio (MediaTek Bluetooth SDIO) driver allows a local attacker with low privileges to permanently hang the Bluetooth stack by triggering an infinite loop in btmtksdio_txrx_work(). The root cause is an inverted timeout logic: a misused time_is_before_jiffies() call, combined via OR, causes the worker loop's exit condition to become unconditionally true after the 5-second timeout fires, so the loop never terminates and never releases the SDIO host. No public exploit has been identified at time of analysis; however, the bug can also trigger organically under hardware stress conditions.
NULL pointer dereference in the Linux kernel Bluetooth HCI connection handler crashes the kernel when aborting a pending connection while `hdev->sent_cmd` is NULL but `req_status` remains `HCI_REQ_PEND`. Local users with Bluetooth access can trigger a general protection fault from the `hci_rx_work()` receive path, producing a kernel panic and denying service to all users on the affected system. No public exploit has been identified and EPSS probability stands at 0.21% (12th percentile), indicating very limited real-world exploitation likelihood.
NULL pointer dereference in the Linux kernel's Bluetooth ISO subsystem can crash the kernel when a race condition is triggered during PA-sync broadcast sink socket operations. The flaw exists in iso_conn_big_sync(), which drops the socket lock to call hci_get_route() and then re-acquires it without re-validating that the connection structure is still intact - allowing concurrent teardown events (iso_disconn_cfm → iso_conn_del → iso_chan_del) to null out iso_pi(sk)->conn before it is dereferenced. A local low-privileged user with access to Bluetooth ISO sockets can reliably crash the kernel, resulting in a denial of service. No public exploit code has been identified and no active exploitation is confirmed.
Device reference count leak in the Linux kernel's firmware_loader subsystem exposes systems to local denial-of-service conditions. Specifically, firmware_upload_register() fails to call put_device() on the fw_dev pointer when alloc_lookup_fw_priv() fails after fw_create_instance() has already called device_initialize(), leaving the embedded struct device reference count unbalanced and potentially preventing proper memory cleanup. Discovered via static analysis and confirmed by manual code review, the issue is patched across multiple stable kernel branches including 6.1.178, 6.6.145, 6.12.96, 6.18.39, and 7.1.4. No public exploit exists and no active exploitation has been observed.
Null pointer dereference in the Linux kernel's cpufreq subsystem during system reboot allows a local attacker with low privileges to crash the kernel. The reboot path (kernel_restart() → device_shutdown() → cpufreq_suspend()) does not invoke freeze_processes() - unlike the normal suspend path - leaving CPU hotplug operations free to run concurrently and free governor_data while cpufreq_suspend() is still reading it, producing a use-after-free null pointer dereference. No public exploit exists and EPSS probability is 0.22% (13th percentile), placing this in low-urgency territory despite the ubiquitous kernel footprint.
Kernel denial-of-service via missing exec_update_lock protection in six Linux procfs functions affects all Linux deployments running kernel 2.6.30 through multiple current stable branches. When a low-privileged local user reads specific /proc entries (stat, map_files, ns symlinks) concurrently with an execve() call in the target process, the missing lock allows ptrace_may_access() to observe transiently invalid credential or memory-map state, triggering a kernel crash. No public exploit identified at time of analysis; EPSS is 0.22% (13th percentile), consistent with low near-term automated exploitation probability.
Resource reference leak in the Linux kernel's POSIX CPU timer subsystem allows a local unprivileged user to exhaust kernel pid references and cause a local denial of service. The flaw exists in `do_cpu_nanosleep()` across all stable kernel branches from 2.6.12 onward: when `posix_cpu_timer_set()` returns an error, the pid reference acquired by the preceding `posix_cpu_timer_create()` call is never released. No public exploit code exists and EPSS sits at 0.22%, consistent with a kernel stability regression fix rather than an actively targeted security flaw.
Denial of service via infinite exception loop in the Linux kernel's s390 architecture affects systems running IBM Secure Execution (Ultravisor) workloads on kernel versions 6.7 through the patched releases. When load_unaligned_zeropad() reads across a page boundary where the adjacent page has been donated to the Ultravisor, the do_secure_storage_access() exception handler cannot resolve the fault and enters an unresolvable loop, permanently hanging the affected execution context. Patched in upstream versions 6.12.97, 7.1.4, and 7.2-rc1 via reversion of DCACHE_WORD_ACCESS support on s390; no public exploit or active exploitation is confirmed at time of analysis.
Uninitialized stack memory use in the Linux kernel HFS and HFS+ filesystem drivers allows a local user to crash the kernel by mounting a crafted filesystem image. Two early-exit paths in `hfs_bnode_read()` - an invalid offset check and a corrected-to-zero length - skip the buffer write, while downstream callers such as `hfs_bnode_read_u16()` and `hfs_bnode_read_u8()` consume the uninitialized stack buffer unconditionally, producing KMSAN uninit-value reports and risking kernel panics. Patches are available across multiple stable kernel branches; no public exploit exists and EPSS is very low at 0.22%.
Hung-task denial-of-service in the Linux kernel's nilfs2 filesystem allows a local low-privileged user to indefinitely stall concurrent filesystem operations by supplying out-of-range segment numbers to the NILFS_IOCTL_CLEAN_SEGMENTS ioctl. Because input validation was deferred deep inside the call chain to nilfs_sufile_updatev() - which emits one printk warning per invalid entry while holding ns_segctor_sem write-locked - repeated invocations saturate the global printk path and keep the semaphore held long enough to block all readers, including routine operations such as chmod(). No public exploit has been identified and EPSS is very low (0.22%, 13th percentile), consistent with a local kernel DoS requiring filesystem access.
Unconditional invocation of cancel_work_sync() in the mtk-jpeg media driver's release path triggers spurious WARN_ON kernel warnings on non-multicore MediaTek JPEG codec platforms where the workqueue callback pointer is NULL. Systems running affected Linux kernel versions (6.6.140-6.18.38, 7.0.4) on MediaTek SoC hardware may experience kernel log pollution or, on kernels compiled with CONFIG_PANIC_ON_WARN=y, a full system denial of service when the JPEG device is released. No public exploit has been identified and no active exploitation is recorded; this is a stability and correctness fix with strictly local trigger conditions.
Integer overflow in the Linux kernel's XFS filesystem exchmaps reservation logic allows a local low-privileged user to crash the system. The bug in xfs_exchmaps_estimate_overhead() adds bmbt and rmapbt overhead to a local resblks variable but then performs the UINT_MAX guard check against the original req->resblks value - the pre-overhead figure - before writing the inflated value back into req->resblks for use in xfs_trans_alloc(). This means the guard intended to prevent integer truncation when the reservation is cast to unsigned int is rendered ineffective, enabling a kernel panic on XFS systems where exchmaps overhead pushes the reservation past UINT_MAX. No public exploit exists and EPSS is 0.21%, placing this firmly in the kernel stability/bug-fix category rather than active threat landscape.
Memory exhaustion in the Linux kernel XFS quota subsystem can be triggered by a local low-privilege user on systems mounting metadir XFS images with user quota enabled. The vulnerability exists in xfs_dqinode_metadir_create(), where failure of xfs_metadir_create() returns without releasing the allocated update and transaction state, and commit failures leave behind caller-owned inode references. Confirmed present from Linux 6.13-rc1 through v7.1.1; patched in stable releases 6.18.39 and 7.1.4.
The Linux kernel BPF verifier incorrectly elides null-check requirements for inner array map lookups when the BPF_F_INNER_MAP flag is set, allowing a local attacker with BPF program loading privileges to trigger a null pointer dereference and crash the kernel. The flaw arises because the verifier uses the inner map template's max_entries for bounds reasoning at verification time, while the runtime helper operates against a concrete inner map that may have a larger max_entries value. A crafted BPF program exploiting the faulty nullness elision can cause a kernel denial of service. No public exploit code has been identified at time of analysis, and EPSS is low at 0.21%.
Out-of-bounds read in the Linux kernel's Samsung Kalmia USB network driver allows a malicious USB device to crash the kernel via a u16 integer underflow in `kalmia_rx_fixup()`. When a crafted bulk-IN frame is 6-11 bytes long, the subtraction `skb->len - 12` underflows to approximately 65,530, bypassing length validation and causing `memcmp()`, `skb_trim()`, and `skb_pull()` to operate past the end of the 14,000-byte RX buffer. No public exploit or active exploitation (CISA KEV) has been identified; EPSS at 0.23% (14th percentile) confirms minimal exploitation interest.
Memory exhaustion in the Linux kernel's USB CDNSP host controller driver allows local users to trigger a kernel-level memory leak by forcing partial allocation failures in `cdnsp_alloc_stream_info()`. Systems running Linux 5.12 through unpatched stable branches accumulate unreleased `stream_ctx_array` allocations that are never freed, eventually causing availability impact on hosts with CDNSP USB hardware. No public exploit exists, EPSS stands at 0.21%, and the vulnerability carries no confidentiality or integrity impact per the CVSS vector.
Incorrect argument passing in dwc3_ulpi_setup() causes the Linux kernel's DWC3 USB controller driver to pass a raw MMIO register pointer (dwc3->regs, typed as void __iomem *) to dwc3_readl() and dwc3_writel(), which instead expect the containing struct dwc3 * pointer. Because the register accessor functions internally dereference the struct to locate ->regs, supplying ->regs directly results in incorrect memory access during ULPI (USB Low Pin Interface) initialization - flagged by the kernel sparse checker as a type-safety violation across address spaces. The practical impact is a potential kernel panic or silent misread/miswrite of a USB control register, manifesting as an availability failure in systems using DWC3 ULPI USB controllers. No public exploit exists and the issue requires local access to trigger.
Kernel denial-of-service in the Linux USB gadget composite driver crashes the system by triggering a KASAN memory fault in the USB_DT_OTG request handler when the configuration list is empty. The flaw exists because list_first_entry() never returns NULL - making the subsequent config NULL-check dead code - so when cdev->configs is empty during a gadget unbind race or when a driver sets is_otg before registering any configuration, composite_setup() dereferences a garbage pointer and memcpy() copies arbitrary kernel memory into the USB response buffer. No public exploit code has been identified, and EPSS is 0.22% (13th percentile), consistent with the local, race-condition nature of the flaw.
Reference count leak in the Linux kernel USB gadget printer driver (f_printer) allows a local user to trigger a permanent kref increment on every rejected concurrent open attempt. Each call to printer_open() that returns -EBUSY still increments dev->kref, but because VFS never invokes ->release() for a failed open, the reference is never balanced. Repeated exploitation exhausts the reference counter, preventing proper driver teardown and causing availability degradation. No public exploit has been identified at time of analysis.
Incomplete error-path cleanup in the Linux kernel's uss720 USB-to-parallel-port driver leaves a parport device registered after probe failure, creating a dangling stale pointer backed by freed private data. Systems where a Belkin F5U002 or similar USS720-based adapter is attached and causes get_1284_register() to fail during uss720_probe() are left with an orphaned kernel parport entry and a reserved parport slot, which can lead to a kernel crash when the stale pp->private_data pointer is later dereferenced. No active exploitation has been identified; the bug was discovered via static-analysis research, and patches are available across all major Linux stable branches.
DMA mapping leak in the Linux kernel's MTU3 USB device controller driver causes local denial-of-service on MediaTek SoC platforms. When mtu3_gadget_queue() maps a USB transfer request via DMA before checking whether the QMU GPD ring has capacity, then mtu3_prepare_transfer() fails and returns -EAGAIN, the error path bypasses the mtu3_req_complete() cleanup helper, permanently stranding the DMA mapping. Repeated exhaustion of the IOMMU aperture can destabilize or crash the kernel. No public exploit code exists and EPSS at 0.22% (13th percentile) confirms negligible exploitation interest; the impact is a local availability-only issue confined to MediaTek USB gadget hardware.
Availability loss in the Linux kernel's USB serial digi_acceleport driver allows a local user to permanently halt data reception on a Digi AccelePort serial port until the USB device is physically reconnected or the driver is rebound. The bug arises from a missing URB resubmission path: when a port is closed while in a throttled state, the read URB is never re-queued, leaving the port silently deaf to incoming data upon next open. No public exploit has been identified and EPSS sits at 0.22% (13th percentile), reflecting the niche hardware dependency and local-only attack surface.
Hard lockup in the Linux kernel's digi_acceleport USB serial driver causes full system freeze when a Digi AccelePort device is physically disconnected during active I/O. The driver's OOB (out-of-band) command path loops indefinitely with interrupts disabled upon persistent URB submission failure, resulting in a kernel hard lockup that effectively crashes the host. Systems with this specific USB-to-serial adapter are affected across a wide range of stable kernel branches from 2.6.12 onward; patches have been released across all active stable series.
Memory leak in the Linux kernel USB ULPI (Universal Serial Bus Link Interface) subsystem allows a local low-privileged user to trigger repeated ULPI device registration failures that leak kernel memory, potentially leading to system instability or denial of service. Affected kernel versions span from 4.2 through multiple pre-patch stable branches, with fixes backported to 5.10.261, 5.15.212, 6.1.178, 6.6.145, 6.12.96, 6.18.39, 7.1.4, and 7.2-rc3. No public exploit or active exploitation has been identified at time of analysis; the EPSS score of 0.22% (13th percentile) reflects minimal probability of weaponization.
NULL pointer dereference in the Linux kernel's usbip virtual UDC (vudc) subsystem crashes the kernel when FunctionFS AIO cancellations are processed. The `vep_alloc_request()` function never initialized `vrequest->udc`, so any AIO cancellation arriving at `vep_dequeue()` - such as one triggered by a host-side `usbip attach` - dereferences a null pointer and produces a kernel oops, denying service to the guest system. A local attacker with low privileges on a system configured for USB/IP with FunctionFS AIO endpoints can reliably reproduce this crash. No public exploit code exists, and EPSS at 0.22% (13th percentile) confirms the path has not seen meaningful exploitation despite the bug reportedly existing for approximately ten years since Linux 4.7.
Kernel memory leak in the Linux USB gadget FunctionFS (f_fs) subsystem allows a local low-privileged user to slowly exhaust kernel memory by repeatedly triggering DMA buffer transfers via the ffs_dmabuf_transfer() code path. Affected kernels span from the commit introducing DMA buffer sharing (7b07a2a7ca02) through versions prior to 6.12.96, 7.1.4, and 7.2-rc3. The root cause is a missing dma_fence_put() call after dma_resv_add_fence(), leaving the initial kref permanently outstanding and causing ffs_dma_fence objects to accumulate without ever being freed. No public exploit has been identified at time of analysis.
The USB gadget FunctionFS (f_fs) subsystem in the Linux kernel fails to initialize the endpoint direction field (epfile->in) until USB host connection, leaving it zero when userspace performs early DMA ioctl calls such as FUNCTIONFS_DMABUF_ATTACH before a host connects. This incorrect DMA direction causes kernel-mode availability loss - crashes or hangs - on systems actively using FunctionFS-based USB gadgets with DMABUF. No public exploit has been identified, and EPSS probability stands at 0.21% (11th percentile), reflecting very low opportunistic exploitation risk.
The block device layer in Linux Kernel hangs indefinitely during NVMe surprise removal when `bdev_mark_dead()` is called with `surprise=true`, because a missing guard causes `sync_blockdev()` to be invoked on an already-dead device, permanently blocking `folio_wait_writeback()`. Triggered via `nvme_reset_work()` I/O-queues-lost path, the bug wedges the NVMe reset worker and every task waiting on it, requiring a system reboot to recover. Discovered by FuzzNvme (syzkaller + FEMU fuzzing framework); no public exploit identified at time of analysis.
The full-scope-disabled client-policy executor in Red Hat Build of Keycloak silently skips enforcement when the fullScopeAllowed attribute is absent from a client registration or update request, allowing a delegated user to create OAuth/OIDC clients with unrestricted scope access. Affected platforms span Red Hat Build of Keycloak, Red Hat Single Sign-On 7, Red Hat Data Grid 8, and JBoss EAP Expansion Pack. An attacker holding delegated registration privileges (CVSS PR:L) can exploit this to issue tokens containing unauthorized role mappings, effectively escalating their effective permissions within any application trusting those tokens. No public exploit code or CISA KEV entry has been identified at time of analysis.
pgAdmin 4 SERVER mode (versions 9.3 through 9.16) allows authenticated users to fully bypass administrator-configured per-tool role-based access controls by exploiting incomplete permission enforcement across backend routes and Socket.IO handlers. The Flask-Security `permissions_required` decorator was applied only to a single front-door route per tool, while all secondary routes and Socket.IO event handlers checked only authentication, not tool permissions - meaning a user who received a 403 on the initial gated endpoint could still drive the complete tool workflow through its other endpoints. Affected tools include Query Tool, Grant Wizard, Schema Diff, ERD, PSQL (interactive terminal), Debugger, Backup, Restore, Maintenance, and Import/Export. No dedicated public exploit has been identified, though the fix commits and GitHub issue are publicly visible; exploitation requires a valid authenticated pgAdmin session.
Denial of service in Apache HttpComponents Client (versions 5.0-alpha1 through 5.6.2) arises when the classic/blocking I/O HttpClient fails to release the underlying connection back to the pool after receiving a response carrying an invalid or unsupported Content-Encoding header. An attacker who controls or can influence the server a client talks to can repeatedly return such headers to leak connections until the pool is exhausted, halting further outbound HTTP requests from the affected Java application. The async I/O HttpClient is explicitly unaffected; there is no public exploit identified at time of analysis and no CVSS or EPSS data was supplied.
SAML HTTP-Redirect binding in Keycloak allows network-based attackers to inject malicious parameters into authentication requests targeting clients configured with wildcard redirect URLs. When a victim user completes authentication, Keycloak appends its legitimate SAML response to the attacker-controlled URL parameters, which can cause susceptible service providers to process the attacker's data instead of the real login assertion - potentially resulting in the victim being logged into the wrong account. No public exploit is identified at time of analysis, and the CVSS score of 3.4 (Low) reflects the high attack complexity and limited integrity-only impact.
Insecure file permissions in ansible-collection-redhat-leapp expose PostgreSQL backup archives to local non-root users on managed RHEL 9 and RHEL 10 nodes. When the `leapp_old_postgresql_data` remediation option is invoked with elevated privileges, the resulting backup archive is written with world-readable or group-readable permissions (CWE-732), allowing any local non-root user on that managed node to extract the full contents of the archived PostgreSQL data. No public exploit has been identified at time of analysis, and exploitation requires local authenticated access to the managed node.
The ansible-collection-redhat-leapp package exposes Ansible controller-side sensitive files to attacker-controlled managed nodes through a manipulated Leapp upgrade report. An attacker who has already obtained privileged write access to a managed node's Leapp report can craft malicious report content that, when an operator executes a specific remediation task, causes the Ansible controller to read and copy its own local files - including private keys and credentials - to the attacker-controlled managed node. No public exploit code has been identified at time of analysis, and this vulnerability is not listed in the CISA KEV catalog; however, the CVSS 3.1 score of 6.2 reflects real-world risk limited by the high privilege prerequisite and required operator interaction.
Uncontrolled decompression in aiohttp's WebSocket client prior to version 3.14.2 allows a malicious server to exhaust CPU and memory on connecting clients by sending RSV1-flagged compressed frames without having negotiated the permessage-deflate extension. The root cause is a missing constructor argument in `client.py:_ws_connect` that caused `WebSocketReader` to default to `compress=True`, violating RFC 6455 §5.2 which mandates failing the connection on undefined reserved bits. This is client-side exposure only - aiohttp server deployments are not affected. No active exploitation has been identified and no public exploit code exists at time of analysis.
Denial of service in Samba's Kerberos KDC password change (kpasswd) service allows authenticated remote attackers to crash the KDC process by submitting specially crafted, malformed ASN.1-encoded password change requests. The server miscalculates the structure size of the incoming request and reads up to six bytes beyond the end of the allocated buffer; when this read strikes unmapped memory, the KDC process terminates rather than simply failing gracefully. No public exploit has been identified and no CISA KEV listing exists, but the disruption of a KDC in an Active Directory or Samba domain environment carries significant operational impact since authentication for the entire domain depends on KDC availability.
Samba's internal DNS server inserts unauthenticated TKEY registration requests into the TKEY name cache before rejecting them, allowing a remote unauthenticated attacker to exhaust the cache by flooding it with arbitrary TKEY names. Eviction of legitimate cache entries blocks TSIG-authenticated DNS query processing, producing a targeted denial of service against Active Directory environments relying on Samba's built-in DNS backend. No public exploit code or CISA KEV listing has been identified at time of analysis; a vendor patch is available from both Samba upstream and Red Hat.
Credential exfiltration in koku-metrics-operator allows a privileged user with edit access to the CostManagementMetricsConfig custom resource to redirect the operator's OAuth token requests to an attacker-controlled endpoint, exposing the tenant's Red Hat SSO client_id and client_secret. The operator's service-account authentication mode unconditionally forwards these credentials to whatever OAuth token URL is configured in the CR, acting as a confused deputy on behalf of the CR editor. No public exploit code has been identified and the vulnerability is not listed in CISA KEV, but the CVSS Scope:Changed metric reflects that the operator process holds credentials inaccessible to the CR editor directly, amplifying impact beyond the initial access tier.
Bearer-token disclosure in the Red Hat koku-metrics-operator lets a user with edit rights over the CostManagementMetricsConfig custom resource redirect metric uploads to an attacker-controlled URL, causing the operator to attach the cluster-global Red Hat Cloud pull-secret token to that outbound request and leak it. Because token authentication is the default (authentication.type=token) and the pull secret is shared cluster-wide, a single privileged tenant can harvest credentials that grant access to the organization's Red Hat Cloud/console.redhat.com services. There is no public exploit identified at time of analysis and it is not listed in CISA KEV; the issue was reported directly by Red Hat's security team.
Stack exhaustion in PHP's phar archive handler allows an attacker to crash the PHP process by supplying a crafted archive containing circular symbolic links, triggering unbounded recursion that depletes the native C stack. Affected versions span PHP 8.2.x before 8.2.33, 8.3.x before 8.3.33, 8.4.x before 8.4.24, and 8.5.x before 8.5.9 across all platforms. Exploitation requires a PHP process to actively parse the malicious archive (UI:A/AV:L); no public exploit has been identified and exploitation likelihood is rated unlikely (E:U in the provided CVSS 4.0 vector), making this a moderate-priority patch for most deployments - with heightened urgency for applications that process user-supplied phar archives.
Uninitialized memory use in Skia, Chrome's 2D graphics rendering engine, exposes process memory contents to remote attackers on Windows via crafted HTML pages in Google Chrome versions prior to 151.0.7922.72. The flaw is strictly an information disclosure issue (CWE-457), with no integrity or availability impact, affecting only the Windows platform per the vulnerability description. No public exploit code has been identified, EPSS places exploitation probability at 0.19% (9th percentile), and SSVC classifies exploitation as none with partial technical impact - collectively suggesting limited real-world risk despite the CVSS C:H rating.
Side-channel GPU information leakage in Google Chrome on Android prior to 151.0.7922.72 enables remote attackers to extract cross-origin data by serving a crafted HTML page. The vulnerability exploits the GPU rendering pipeline to observe timing or state artifacts that escape the Same-Origin Policy boundary, allowing partial disclosure of data from other origins loaded in the browser. No active exploitation has been confirmed (not in CISA KEV), and the EPSS score of 0.17% (7th percentile) reflects low observed exploitation probability; a vendor-released patch is available.
Same-origin policy bypass in Google Chrome's DevTools component prior to version 151.0.7922.72 allows an unauthenticated remote attacker to perform unauthorized cross-origin write operations against a victim's browser context. Exploitation requires the victim to visit a specially crafted HTML page (UI:R), making it a socially-engineered threat rather than a purely opportunistic one. EPSS is low at 0.22% (13th percentile) and no CISA KEV listing or public exploit code has been identified, consistent with the Chromium team's own 'Medium' severity rating; however, Chrome's ubiquitous deployment makes even medium-severity findings broadly relevant.
Cross-origin data leakage in Google Chrome's WebProtect component (prior to 151.0.7922.72) enables a remote attacker who has already compromised the renderer process to exfiltrate cross-origin data by loading a crafted HTML page. The flaw stems from an inappropriate implementation of origin validation (CWE-346), allowing WebProtect to fail its primary function of enforcing same-origin boundaries within Chrome's sandboxed renderer. No active exploitation has been confirmed by CISA KEV, no public POC has been identified, and EPSS at 0.21% (11th percentile) indicates low near-term exploitation probability - though this vulnerability's relevance as a component in chained browser exploit chains elevates its strategic significance beyond the raw score.
Cross-origin data leakage in Google Chrome's GPU implementation prior to version 151.0.7922.72 allows a remote attacker who has already compromised the renderer process to read cross-origin rendered content via a crafted HTML page. The flaw stems from an origin validation error (CWE-346) in the GPU compositor layer, enabling sandbox-escaped renderer processes to access pixel or texture data belonging to other security origins. EPSS at 0.21% (11th percentile) and no CISA KEV listing indicate low current exploitation probability; no public exploit code has been identified at time of analysis.
Cross-origin data leakage in Google Chrome's NoStatePrefetch feature exposes confidential resources across origin boundaries when a victim visits an attacker-crafted HTML page, affecting all Chrome versions prior to 151.0.7922.72. The flaw exploits observable side-channel signals produced during background prefetch operations to infer cross-origin state, violating Same-Origin Policy protections. No public exploit or active exploitation has been identified; EPSS sits at 0.21% (11th percentile), indicating low near-term exploitation likelihood despite the product's ubiquitous deployment footprint.
Cross-origin data leakage in Google Chrome's Skia graphics engine affects all Chrome versions prior to 151.0.7922.72, enabling remote attackers to read sensitive data across browser security boundaries when a victim visits a specially crafted HTML page. The root cause is uninitialized memory use (CWE-457) within the Skia 2D rendering library, which may expose residual memory contents - potentially including fragments from other origins - violating the Same-Origin Policy at the memory level. No public exploit code has been identified and EPSS sits at 0.23% (14th percentile), indicating low exploitation probability at time of analysis despite the vulnerability being remotely triggerable without authentication.
Navigation restriction bypass in Google Chrome's Presentation component (all versions prior to 151.0.7922.72) enables a remote attacker to circumvent enforced browser navigation policies via a specially crafted HTML page. The flaw carries a high integrity impact (I:H) with no privilege requirement on the attacker's side, though victim interaction is mandatory. EPSS sits at 0.22% (13th percentile) and SSVC classifies exploitation as 'none' with a non-automatable profile, placing this in the patch-but-not-emergency tier for most organizations. No public exploit code identified at time of analysis.
UI spoofing in Google Chrome's Extensions component enables an attacker who persuades a victim to install a malicious extension to misrepresent browser security indicators, affecting all Chrome desktop versions prior to 151.0.7922.72. The flaw (CWE-451) stems from incorrect rendering of security UI elements within the Extensions subsystem, granting a crafted extension the ability to deceive users about their browser's security state or extension permissions. No public exploit identified at time of analysis, and EPSS at 0.13% (3rd percentile) reflects very low exploitation probability; this CVE is not listed in CISA KEV.
Same-origin policy bypass in Google Chrome's Blink rendering engine (versions prior to 151.0.7922.72) allows a remote attacker to violate cross-origin isolation boundaries by luring a user to visit a specially crafted HTML page. The CVSS vector assigns High integrity impact with no privilege requirement, meaning successful exploitation could enable an attacker to inject or manipulate cross-origin content within the victim's browser session. No public exploit has been identified at time of analysis and EPSS sits at 0.22% (13th percentile), with SSVC confirming no observed exploitation - placing this in the patch-on-schedule rather than emergency-response tier.
Cross-origin data leakage in Google Chrome's Autofill subsystem (all versions prior to 151.0.7922.72) enables a remote attacker to read sensitive form data across origins by serving a crafted HTML page to the victim. The root cause is an origin validation error (CWE-346) where the Autofill implementation fails to enforce proper origin boundaries, violating the browser same-origin policy for autofill-populated fields. No public exploit code has been identified and the EPSS score of 0.21% places this in the 11th percentile, indicating very low current exploitation probability; a vendor patch is available.
Cross-origin data leakage in Google Chrome's Payments subsystem prior to version 151.0.7922.72 allows a remote attacker to exfiltrate limited cross-origin information by luring a victim to a crafted HTML page. The root cause is insufficient policy enforcement (CWE-346) within the Payments component, meaning Chrome fails to properly validate origin boundaries when handling payment-related operations. No public exploit code has been identified and EPSS is very low (0.21%), but the ubiquitous deployment of Chrome elevates aggregate exposure.
Cross-origin data leakage in Google Chrome's ANGLE graphics component (versions prior to 151.0.7922.72) exposes residual memory contents to remote attackers via a crafted HTML page. ANGLE - Chrome's graphics abstraction layer - fails to initialize memory before use (CWE-457), allowing pixel or buffer data from one origin to be read by another. No public exploit identified at time of analysis; EPSS probability is 0.23% (14th percentile), indicating low current exploitation activity despite Chrome's massive deployment footprint.
Side-channel information leakage in Google Chrome's Autofill feature (versions prior to 151.0.7922.72) enables cross-origin data exfiltration by exploiting observable behavioral differences during Autofill interactions. An unauthenticated remote attacker who delivers a crafted HTML page and persuades the victim to perform specific UI gestures can infer data belonging to other origins, bypassing Same-Origin Policy protections. EPSS is very low at 0.21% (11th percentile) and no active exploitation is confirmed via CISA KEV, making this a real but modest real-world risk - primarily meaningful where sensitive Autofill data (saved addresses, payment details) is stored in the target browser profile.
Cross-origin data leakage in Google Chrome's Actor component (all versions prior to 151.0.7922.72) allows a remote attacker who has already compromised the renderer process to exfiltrate sensitive cross-origin data via a crafted HTML page. This is a secondary-stage information disclosure primitive - it requires both user interaction and a pre-compromised renderer, making it a chained exploit component rather than a standalone entry point. No public exploit or active exploitation has been identified; EPSS sits at 0.29% (21st percentile), consistent with low real-world exploitation likelihood at this time.
Side-channel information leakage in Google Chrome's Media component (versions prior to 151.0.7922.72) enables remote attackers to exfiltrate cross-origin data by luring a victim to a crafted HTML page that exploits observable timing or state signals in media processing. This bypasses the browser's Same-Origin Policy protections, allowing an attacker's origin to infer sensitive data loaded from a different origin. No public exploit has been identified at time of analysis, and the EPSS score of 0.29% (21st percentile) reflects low current exploitation probability; a vendor-released patch is available.
Cross-origin data leakage in Google Chrome's FileSystem component prior to version 151.0.7922.72 enables a remote attacker to read data across origin boundaries by luring a victim to a crafted HTML page. The root cause is insufficient policy enforcement (CWE-346 Origin Validation Error) in Chrome's FileSystem API, which fails to enforce same-origin policy checks correctly, exposing data that should be sandboxed to its originating origin. No public exploit code has been identified at time of analysis, and the EPSS score of 0.29% (21st percentile) reflects low current exploitation likelihood; however, the vendor patch is available and should be applied promptly given Chrome's ubiquitous deployment footprint.
Uninitialized memory use in ANGLE, Chrome's graphics translation layer, exposes cross-origin data to remote attackers who can direct users to a crafted HTML page. Affected are all Google Chrome desktop releases prior to 151.0.7922.72 across supported platforms. No public exploit code or active exploitation has been identified; EPSS probability is low at 0.31% (24th percentile), and the impact is bounded to partial confidentiality loss with no integrity or availability consequences.
The BGR decoder in ImageMagick prior to 7.1.2-27 fails to check for end-of-file at all necessary code paths, allowing remote unauthenticated attackers to trigger a heap buffer over-read via a crafted image file. The CVSS vector (AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N) confirms network-exploitable, zero-prerequisite access with impact limited to partial memory disclosure. No active exploitation (CISA KEV) and no public POC have been identified at the time of analysis; a vendor-released patch is available in version 7.1.2-27.
CRLF injection in undici's HTTP/1.1 dispatcher enables HTTP header injection and potential request smuggling when applications pass duck-typed blob-like bodies whose type property is populated from untrusted input. Affected are the 6.x branch before 6.28.0, all 7.x releases from 7.0.0 through 7.28.x, and all 8.x releases from 8.0.0 through 8.8.x, when callers use the request, stream, pipeline, or dispatch APIs with hand-rolled blob-shaped objects rather than native Blob instances or the fetch API. An attacker who controls the type value can append arbitrary HTTP headers to the outgoing request, potentially smuggling a second request past an upstream proxy or server. No public exploit identified at time of analysis and KEV status is not confirmed in the provided data.
Response body/Content-Length desynchronization in undici's retry interceptor allows a malicious or faulty upstream server to induce HTTP response smuggling against downstream clients of proxy and gateway applications. Affected versions span undici before 6.28.0, the 7.x line before 7.29.0, and the 8.x line before 8.9.0. An upstream that deliberately returns a partial response with mismatched framing and closes the socket early can cause the retry interceptor to assemble a body of a different length than the attached Content-Length, leading downstream consumers to experience response desynchronization, connection hangs, or data corruption. No public exploit has been identified at time of analysis, and this is not listed in CISA KEV.
Server-Side Request Forgery in Pydantic AI's UI adapter layer (versions 1.65.0-1.105.0 and 2.0.0b1-2.0.0b5) allows unauthenticated network clients to redirect the application server into fetching arbitrary cloud-storage objects using the server's own credentials. UploadedFile references - carrying provider file IDs or cloud-storage URIs such as s3:// or gs:// - submitted via client message history bypass the scheme allowlist applied to ordinary file URLs and are resolved server-side using the application's IAM role, service account, or API key. No public exploit has been identified at time of analysis, and no CISA KEV listing exists, but the scope-changing confidentiality impact (CVSS S:C/C:H) warrants prompt patching for any deployment exposing the Vercel AI adapter or similar UI adapters to untrusted clients.
Tool call injection in Pydantic AI's UI adapters (AG-UI via Agent.to_ag_ui()/AGUIAdapter, and Vercel AI via VercelAIAdapter) enables remote unauthenticated clients to force execution of registered server-side tools with attacker-supplied arguments, bypassing the framework's message-sanitization defense. The flaw affects pydantic-ai and pydantic-ai-slim versions 1.88.0-1.107.0 and 2.0.0b1-2.4.x. No public exploit code or CISA KEV listing exists at time of analysis, but the attack is low-complexity and exploitable against default configurations of applications using either UI adapter.
Arbitrary file overwrite in Develar app-builder (all versions ≤ 4.2.0 on macOS) allows an unauthenticated remote attacker to escape the ZIP extraction directory by exploiting APFS Unicode case-folding combined with symlink following in the `zipx.Unzip` routine. A crafted archive containing a symlink entry `ss` pointing to an external victim file and a subsequent regular-file entry `ß` causes the builder to follow the symlink - because APFS folds `ß` to `ss` - overwriting the target with attacker-controlled content. SSVC rates the attack automatable with no confirmed exploitation; no public exploit code has been identified at time of analysis, though CERT/CC advisory VU#293714 and ENISA EUVD-2026-50423 confirm coordinated disclosure.
Cookie attribute injection in undici's setCookie function allows attacker-controlled domain values or unparsed array entries to silently append, strip, or override attributes such as SameSite, HttpOnly, and Secure. Affected are all undici releases before 6.28.0, the 7.x branch before 7.29.0, and the 8.x branch before 8.9.0. In multi-tenant or reverse-proxy deployments that scope session cookies to tenant-supplied domains, this flaw enables CSRF protection bypass by downgrading or overriding the SameSite attribute, with no public exploit identified at time of analysis.
Node.js Permission Model fails to enforce filesystem write restrictions for the trace events subsystem, allowing a local attacker with low privileges to write files to paths explicitly blocked by the operator via the --experimental-permission flag. Rated Low severity (CVSS 3.3), this flaw affects Node.js 22.x through 22.23.1, 24.x through 24.18.0, and 26.x through 26.5.0, and is one of ten CVEs addressed in the July 2026 Node.js security release. No public exploit code has been identified and the vulnerability is not listed in CISA KEV; the fix is available in Node.js 26.5.1 released 2026-07-29.
IP-based security controls in applications using @fastify/forwarded before 3.0.2 can be bypassed by injecting a horizontal tab character into X-Forwarded-For header entries. The parser strips only space characters during whitespace normalization, violating RFC 7230's definition of optional whitespace (which includes both SP and HTAB), causing the resolved IP string to carry a trailing tab. Any application performing exact-string security decisions on the resolved address - allowlists, blocklists, per-IP rate limit keys, or audit log correlation - will silently fail to match the expected value, allowing an attacker to evade those controls. No public exploit has been identified at time of analysis, and no KEV listing exists; however, exploitation requires no authentication or special configuration beyond controlling HTTP request headers.
Non-atomic JSON Patch application in cJSON through 1.7.19 allows an attacker who controls the patch document to silently corrupt the target JSON structure even when the API returns a failure status. The apply_patch() function in cJSON_Utils.c deletes existing target members before completing operation validation, meaning a malformed replace (missing value) or move (unresolvable destination) permanently destroys data from the in-memory document tree - defeating the RFC 6902 all-or-nothing contract that callers rely on to reject bad patches safely. No public exploit has been identified at time of analysis, but a detailed researcher disclosure and direct source code reference are publicly available, which may lower the barrier to exploitation; KEV status is not confirmed.
Path traversal in open-iscsi allows remote MITM attackers to create root-owned files at arbitrary filesystem locations and inject lines into records by embedding RFC 3722-prohibited characters in IQN target names returned during iSNS or firmware discovery. All open-iscsi versions through commit 668ca1df9c9a1e9bdd5c999ae1d67c9c8909237e are affected across the iSNS discovery, firmware discovery, and command-line target-name paths. No public exploit code or CISA KEV listing has been identified at time of analysis; however, the CVSS 4.0 score of 6.9 with network vector and no privileges required warrants attention in enterprise storage environments where iSCSI discovery is exposed.
Terminal control sequence injection in diff-so-fancy exposes developers to output manipulation, filename spoofing, terminal screen clearing, and clipboard hijacking when viewing diffs containing attacker-controlled content. The application filters only ANSI SGR sequences while passing OSC, CSI, carriage returns, and other VT100 control characters through unsanitized to the terminal, affecting all versions prior to commit 9c81294. No public exploit is identified at time of analysis, but the low attack complexity (CVSS 4.0 AV:L/AC:L) means the primary barrier is only the requirement for a developer to actively view the malicious diff.
Unauthenticated disabling of the ACME service in Dogtag PKI allows persistent denial of service via URL manipulation. By appending a trailing slash to admin-only enable/disable endpoints, an attacker can bypass the Tomcat authentication constraint while RESTEasy routes the request, enabling unauthorized toggling. No public exploit or active exploitation has been reported.
Resource exhaustion in Ninenines Cowboy's HTTP/1.1 handler allows unauthenticated remote attackers to cause denial of service by sending many header lines with the same name, bypassing the max_headers limit due to counting distinct names instead of total lines. Affects Cowboy versions 2.0.0-pre.4 to 2.17.x; fixed in 2.18.0. No public exploit or active exploitation has been reported at this time.
Safari and WebKit-based apps across Apple platforms crash when a user visits a specially crafted website, due to a state management flaw (CWE-400: Uncontrolled Resource Consumption) in versions prior to 26.6. Affected platforms include Safari on macOS Tahoe, iOS 26, iPadOS 26, and visionOS 26. No confidentiality or integrity impact is present - the consequence is an app-level denial-of-service. No public exploit code exists and CISA SSVC confirms no active exploitation, with EPSS at 0.16% (6th percentile), indicating very low exploitation likelihood at time of analysis.
UI spoofing in Apple Safari 26.5 and earlier allows attackers using cross-origin framing to misrepresent website identity. The vulnerability requires user interaction (visiting a malicious page) and could lead to phishing or deceptive actions. No active exploitation or public exploit code identified; patch is available.
Bypass of iframe sandboxing policy in Apple Safari and all Apple operating systems through version 26.5 allows malicious web content to violate sandbox restrictions. Affected users include anyone running Safari on iOS, iPadOS, macOS Tahoe, tvOS, visionOS, or watchOS prior to version 26.6. An attacker can cause integrity violations, such as injecting or modifying content within sandboxed iframes, but requires user interaction (clicking a link or visiting a malicious page) and no public exploit or active exploitation has been identified at this time.
An integer overflow in GIMP's file-sgi plugin, triggered by a crafted RLE-compressed SGI image, causes denial of service by crashing the plugin. The vulnerability affects GIMP versions with the file-sgi plugin and requires a user to open a malicious image file. No public exploit or active exploitation is known at this time; the issue is being tracked by GNOME and Red Hat.
Path traversal vulnerability in Erlang OTP's stdlib zip module allows attackers to write files outside the intended extraction directory by exploiting flawed validation of relative paths in zip archives. Affected are OTP versions 27.1 before 27.3.4.15, 28.5.0.4, and 29.0.4. Successful exploitation requires user interaction (extracting a crafted zip) and has limited impact on integrity and availability; no active exploitation has been reported.
Remote denial of service in llama.cpp allows unauthenticated attackers to exhaust server resources via crafted JSON schema inputs that trigger unbounded recursion in the JSON-schema-to-GBNF conversion routine. Affected versions include commit e15efe0 and prior. A fix exists in a pending pull request; no active exploitation reported.
Denial of service in SQLite 3.41 due to a use-after-free in the JSON extraction function. A local unauthenticated attacker can trigger a crash by supplying a malicious JSON input, causing the host application to terminate. No active exploitation has been confirmed in the wild, but public exploit proof-of-concept may exist.
Memory exhaustion in ImageMagick's magick CLI tool allows a local attacker to cause denial of service by repeatedly supplying malformed command-line arguments. Affects versions before 7.1.2-27. Vendor patch is available; no public exploit or active exploitation reported.
NULL pointer dereference in the Linux kernel tracing subsystem (`func_set_flag()`) allows a local attacker with tracefs write access to trigger a kernel panic and crash the system. The flaw arises when a process holds an open file descriptor to a function tracer option (e.g., `func_stack_trace`) while a concurrent tracer switch sets `tr->current_trace_flags` to NULL; a subsequent write to that option file then dereferences the NULL pointer before the function tracer guard executes. No active exploitation (KEV) or public exploit code has been identified; EPSS is 0.20% at the 10th percentile, consistent with a localized, technically specific DoS condition.
Kernel crash (availability loss) in the Linux gpio/eic-sprd Spreadtrum EIC GPIO driver affects systems running PREEMPT_RT kernels with Unisoc/Spreadtrum SoCs. The IRQ startup path - request_threaded_irq() → irq_startup() → sprd_eic_irq_unmask() - calls sprd_eic_update(), which acquires a regular spinlock_t via spin_lock_irqsave(). On PREEMPT_RT kernels, spinlock_t is a sleeping lock; sleeping in a non-sleepable IRQ-disabled context triggers a kernel BUG: 'sleeping function called from invalid context', crashing the system. No public exploit has been identified at time of analysis; patches are available across all maintained stable branches.
Kernel panic risk in the Linux gpio-sch driver exposes PREEMPT_RT systems with legacy Intel SCH GPIO hardware to a local denial of service. The SCH IRQ unmask callback acquires a regular spinlock inside a non-sleepable IRQ-setup context; under PREEMPT_RT, that spinlock is reimplemented as a sleeping lock, triggering a lockdep BUG and potential kernel panic. Patches are available across all active stable branches (5.15.212, 6.1.178, 6.6.145, 6.12.97, 6.18.39, 7.1.4); no public exploit exists and EPSS is 0.21% (percentile 12%).
NULL pointer dereference in the Linux kernel's hid_logitech_dj driver causes a kernel crash (DoS) when a crafted HID descriptor is processed during device probe. The flaw exists in logi_dj_probe(): when a Logitech Unifying receiver emulation presents report ID 0x20 with only padding output items, hid-core registers the report with rep->maxfield == 0. The error-path validation then dereferences rep->field[0]->report_count while printing a diagnostic message, triggering a KASAN-visible null-ptr-deref and kernel panic. No public exploit is identified at time of analysis, but the kernel commit descriptions include a full uhid-based reproduction method, lowering the barrier for local attackers.
Resource exhaustion via file reference leak in the Linux kernel's io_uring NOP opcode handler allows a local, low-privileged user to gradually exhaust kernel memory. The flaw occurs when a NOP submission carries the generic IOSQE_FIXED_FILE sqe flag (setting REQ_F_FIXED_FILE) but omits the NOP-specific IORING_NOP_FIXED_FILE flag in nop_flags: io_nop() acquires a real file reference via io_file_get_normal(), yet io_put_file() skips releasing it because REQ_F_FIXED_FILE is set, producing one leaked 176-byte kernel file object per operation. No public exploit has been identified at time of analysis, and the vulnerability is not listed in the CISA KEV catalog.
Availability impact in the Linux kernel's io_uring work queue (io-wq) subsystem allows a local, low-privileged user to cause an extended hang during io_uring ring closure. The root cause is a race condition in io_worker_handle_work(): the IO_WQ_BIT_EXIT check is snapshotted once before processing a linked-work chain, so if io_wq_exit_start() fires mid-chain, remaining linked items execute with a stale do_kill=false, skip IO_WQ_WORK_CANCEL, and continue running after exit has been initiated. No public exploit is identified at time of analysis; the issue was surfaced by syzbot using slow MSR-device reads with linked work chains and immediate ring close.
Kernel availability loss in the Linux mm/shrinker subsystem allows a local user with debugfs access to trigger a kernel BUG by reading the debugfs 'count' file of a memcg-aware shrinker such as zswap. The shrinker_debugfs_count_show() function incorrectly holds rcu_read_lock() while invoking ->count_objects() callbacks; when zswap's callback attempts to flush memcg stats via css_rstat_flush(), it sleeps inside an RCU read-side critical section - a fatal kernel invariant violation. No public exploit has been identified at time of analysis, and EPSS is very low at 0.21% (12th percentile).
Kernel panic via NULL pointer dereference in the Linux kernel's memory shrinker debugfs interface allows a local attacker to crash the system. The shrinker_debugfs_add() function unconditionally creates 'count' and 'scan' debugfs files for all registered shrinkers, but some shrinkers - notably the xen-backend shrinker - only implement one of the two required callbacks (count_objects or scan_objects). Writing to the exposed 'scan' debugfs file on such a shrinker dispatches through a NULL function pointer, triggering a kernel panic. No public exploit has been identified at time of analysis, and EPSS score of 0.21% reflects low exploitation probability. Patches are available across multiple stable kernel branches.
NULL pointer dereference in the Linux kernel's swap cgroup subsystem (`mm/swap_cgroup.c`) causes a kernel panic and full system crash on swapless hosts running kernel versions after commit `bea67dcc5eea`. The batched PTE zap path introduced by that commit calls `lookup_swap_cgroup_id()` without validating the swap entry against `swap_info[]`, so any process exit involving a PTE corrupted into a type-0 swap entry dereferences the uninitialized `ctrl->map` pointer and takes the host down. The crash pattern was confirmed in production on kernel 6.12.58 and independently reproduced by syzbot; no public exploit exists, but the kernel teardown path is deterministically vulnerable once the prerequisite conditions are present.
Softlockup in the Linux kernel's swap memory reclaim path allows a local user to trigger a kernel panic by generating heavy swap pressure on systems with large swap devices. The swap_reclaim_full_clusters() function, called from a kworker context, iterates through potentially thousands of full swap clusters without yielding the CPU via cond_resched(), starving the scheduler and eventually causing the kernel watchdog timer to fire. The vulnerability was confirmed in a real-world stress test on a 320-CPU arm64 machine with ~1TB RAM and an 8.6GB swap device, resulting in a kernel panic. No public exploit code exists and EPSS is very low (0.21%, 11th percentile).
Denial of service in the Linux kernel btmtksdio (MediaTek Bluetooth SDIO) driver allows a local attacker with low privileges to permanently hang the Bluetooth stack by triggering an infinite loop in btmtksdio_txrx_work(). The root cause is an inverted timeout logic: a misused time_is_before_jiffies() call, combined via OR, causes the worker loop's exit condition to become unconditionally true after the 5-second timeout fires, so the loop never terminates and never releases the SDIO host. No public exploit has been identified at time of analysis; however, the bug can also trigger organically under hardware stress conditions.
NULL pointer dereference in the Linux kernel Bluetooth HCI connection handler crashes the kernel when aborting a pending connection while `hdev->sent_cmd` is NULL but `req_status` remains `HCI_REQ_PEND`. Local users with Bluetooth access can trigger a general protection fault from the `hci_rx_work()` receive path, producing a kernel panic and denying service to all users on the affected system. No public exploit has been identified and EPSS probability stands at 0.21% (12th percentile), indicating very limited real-world exploitation likelihood.
NULL pointer dereference in the Linux kernel's Bluetooth ISO subsystem can crash the kernel when a race condition is triggered during PA-sync broadcast sink socket operations. The flaw exists in iso_conn_big_sync(), which drops the socket lock to call hci_get_route() and then re-acquires it without re-validating that the connection structure is still intact - allowing concurrent teardown events (iso_disconn_cfm → iso_conn_del → iso_chan_del) to null out iso_pi(sk)->conn before it is dereferenced. A local low-privileged user with access to Bluetooth ISO sockets can reliably crash the kernel, resulting in a denial of service. No public exploit code has been identified and no active exploitation is confirmed.
Device reference count leak in the Linux kernel's firmware_loader subsystem exposes systems to local denial-of-service conditions. Specifically, firmware_upload_register() fails to call put_device() on the fw_dev pointer when alloc_lookup_fw_priv() fails after fw_create_instance() has already called device_initialize(), leaving the embedded struct device reference count unbalanced and potentially preventing proper memory cleanup. Discovered via static analysis and confirmed by manual code review, the issue is patched across multiple stable kernel branches including 6.1.178, 6.6.145, 6.12.96, 6.18.39, and 7.1.4. No public exploit exists and no active exploitation has been observed.
Null pointer dereference in the Linux kernel's cpufreq subsystem during system reboot allows a local attacker with low privileges to crash the kernel. The reboot path (kernel_restart() → device_shutdown() → cpufreq_suspend()) does not invoke freeze_processes() - unlike the normal suspend path - leaving CPU hotplug operations free to run concurrently and free governor_data while cpufreq_suspend() is still reading it, producing a use-after-free null pointer dereference. No public exploit exists and EPSS probability is 0.22% (13th percentile), placing this in low-urgency territory despite the ubiquitous kernel footprint.
Kernel denial-of-service via missing exec_update_lock protection in six Linux procfs functions affects all Linux deployments running kernel 2.6.30 through multiple current stable branches. When a low-privileged local user reads specific /proc entries (stat, map_files, ns symlinks) concurrently with an execve() call in the target process, the missing lock allows ptrace_may_access() to observe transiently invalid credential or memory-map state, triggering a kernel crash. No public exploit identified at time of analysis; EPSS is 0.22% (13th percentile), consistent with low near-term automated exploitation probability.
Resource reference leak in the Linux kernel's POSIX CPU timer subsystem allows a local unprivileged user to exhaust kernel pid references and cause a local denial of service. The flaw exists in `do_cpu_nanosleep()` across all stable kernel branches from 2.6.12 onward: when `posix_cpu_timer_set()` returns an error, the pid reference acquired by the preceding `posix_cpu_timer_create()` call is never released. No public exploit code exists and EPSS sits at 0.22%, consistent with a kernel stability regression fix rather than an actively targeted security flaw.
Denial of service via infinite exception loop in the Linux kernel's s390 architecture affects systems running IBM Secure Execution (Ultravisor) workloads on kernel versions 6.7 through the patched releases. When load_unaligned_zeropad() reads across a page boundary where the adjacent page has been donated to the Ultravisor, the do_secure_storage_access() exception handler cannot resolve the fault and enters an unresolvable loop, permanently hanging the affected execution context. Patched in upstream versions 6.12.97, 7.1.4, and 7.2-rc1 via reversion of DCACHE_WORD_ACCESS support on s390; no public exploit or active exploitation is confirmed at time of analysis.
Uninitialized stack memory use in the Linux kernel HFS and HFS+ filesystem drivers allows a local user to crash the kernel by mounting a crafted filesystem image. Two early-exit paths in `hfs_bnode_read()` - an invalid offset check and a corrected-to-zero length - skip the buffer write, while downstream callers such as `hfs_bnode_read_u16()` and `hfs_bnode_read_u8()` consume the uninitialized stack buffer unconditionally, producing KMSAN uninit-value reports and risking kernel panics. Patches are available across multiple stable kernel branches; no public exploit exists and EPSS is very low at 0.22%.
Hung-task denial-of-service in the Linux kernel's nilfs2 filesystem allows a local low-privileged user to indefinitely stall concurrent filesystem operations by supplying out-of-range segment numbers to the NILFS_IOCTL_CLEAN_SEGMENTS ioctl. Because input validation was deferred deep inside the call chain to nilfs_sufile_updatev() - which emits one printk warning per invalid entry while holding ns_segctor_sem write-locked - repeated invocations saturate the global printk path and keep the semaphore held long enough to block all readers, including routine operations such as chmod(). No public exploit has been identified and EPSS is very low (0.22%, 13th percentile), consistent with a local kernel DoS requiring filesystem access.
Unconditional invocation of cancel_work_sync() in the mtk-jpeg media driver's release path triggers spurious WARN_ON kernel warnings on non-multicore MediaTek JPEG codec platforms where the workqueue callback pointer is NULL. Systems running affected Linux kernel versions (6.6.140-6.18.38, 7.0.4) on MediaTek SoC hardware may experience kernel log pollution or, on kernels compiled with CONFIG_PANIC_ON_WARN=y, a full system denial of service when the JPEG device is released. No public exploit has been identified and no active exploitation is recorded; this is a stability and correctness fix with strictly local trigger conditions.
Integer overflow in the Linux kernel's XFS filesystem exchmaps reservation logic allows a local low-privileged user to crash the system. The bug in xfs_exchmaps_estimate_overhead() adds bmbt and rmapbt overhead to a local resblks variable but then performs the UINT_MAX guard check against the original req->resblks value - the pre-overhead figure - before writing the inflated value back into req->resblks for use in xfs_trans_alloc(). This means the guard intended to prevent integer truncation when the reservation is cast to unsigned int is rendered ineffective, enabling a kernel panic on XFS systems where exchmaps overhead pushes the reservation past UINT_MAX. No public exploit exists and EPSS is 0.21%, placing this firmly in the kernel stability/bug-fix category rather than active threat landscape.
Memory exhaustion in the Linux kernel XFS quota subsystem can be triggered by a local low-privilege user on systems mounting metadir XFS images with user quota enabled. The vulnerability exists in xfs_dqinode_metadir_create(), where failure of xfs_metadir_create() returns without releasing the allocated update and transaction state, and commit failures leave behind caller-owned inode references. Confirmed present from Linux 6.13-rc1 through v7.1.1; patched in stable releases 6.18.39 and 7.1.4.
The Linux kernel BPF verifier incorrectly elides null-check requirements for inner array map lookups when the BPF_F_INNER_MAP flag is set, allowing a local attacker with BPF program loading privileges to trigger a null pointer dereference and crash the kernel. The flaw arises because the verifier uses the inner map template's max_entries for bounds reasoning at verification time, while the runtime helper operates against a concrete inner map that may have a larger max_entries value. A crafted BPF program exploiting the faulty nullness elision can cause a kernel denial of service. No public exploit code has been identified at time of analysis, and EPSS is low at 0.21%.
Out-of-bounds read in the Linux kernel's Samsung Kalmia USB network driver allows a malicious USB device to crash the kernel via a u16 integer underflow in `kalmia_rx_fixup()`. When a crafted bulk-IN frame is 6-11 bytes long, the subtraction `skb->len - 12` underflows to approximately 65,530, bypassing length validation and causing `memcmp()`, `skb_trim()`, and `skb_pull()` to operate past the end of the 14,000-byte RX buffer. No public exploit or active exploitation (CISA KEV) has been identified; EPSS at 0.23% (14th percentile) confirms minimal exploitation interest.
Memory exhaustion in the Linux kernel's USB CDNSP host controller driver allows local users to trigger a kernel-level memory leak by forcing partial allocation failures in `cdnsp_alloc_stream_info()`. Systems running Linux 5.12 through unpatched stable branches accumulate unreleased `stream_ctx_array` allocations that are never freed, eventually causing availability impact on hosts with CDNSP USB hardware. No public exploit exists, EPSS stands at 0.21%, and the vulnerability carries no confidentiality or integrity impact per the CVSS vector.
Incorrect argument passing in dwc3_ulpi_setup() causes the Linux kernel's DWC3 USB controller driver to pass a raw MMIO register pointer (dwc3->regs, typed as void __iomem *) to dwc3_readl() and dwc3_writel(), which instead expect the containing struct dwc3 * pointer. Because the register accessor functions internally dereference the struct to locate ->regs, supplying ->regs directly results in incorrect memory access during ULPI (USB Low Pin Interface) initialization - flagged by the kernel sparse checker as a type-safety violation across address spaces. The practical impact is a potential kernel panic or silent misread/miswrite of a USB control register, manifesting as an availability failure in systems using DWC3 ULPI USB controllers. No public exploit exists and the issue requires local access to trigger.
Kernel denial-of-service in the Linux USB gadget composite driver crashes the system by triggering a KASAN memory fault in the USB_DT_OTG request handler when the configuration list is empty. The flaw exists because list_first_entry() never returns NULL - making the subsequent config NULL-check dead code - so when cdev->configs is empty during a gadget unbind race or when a driver sets is_otg before registering any configuration, composite_setup() dereferences a garbage pointer and memcpy() copies arbitrary kernel memory into the USB response buffer. No public exploit code has been identified, and EPSS is 0.22% (13th percentile), consistent with the local, race-condition nature of the flaw.
Reference count leak in the Linux kernel USB gadget printer driver (f_printer) allows a local user to trigger a permanent kref increment on every rejected concurrent open attempt. Each call to printer_open() that returns -EBUSY still increments dev->kref, but because VFS never invokes ->release() for a failed open, the reference is never balanced. Repeated exploitation exhausts the reference counter, preventing proper driver teardown and causing availability degradation. No public exploit has been identified at time of analysis.
Incomplete error-path cleanup in the Linux kernel's uss720 USB-to-parallel-port driver leaves a parport device registered after probe failure, creating a dangling stale pointer backed by freed private data. Systems where a Belkin F5U002 or similar USS720-based adapter is attached and causes get_1284_register() to fail during uss720_probe() are left with an orphaned kernel parport entry and a reserved parport slot, which can lead to a kernel crash when the stale pp->private_data pointer is later dereferenced. No active exploitation has been identified; the bug was discovered via static-analysis research, and patches are available across all major Linux stable branches.
DMA mapping leak in the Linux kernel's MTU3 USB device controller driver causes local denial-of-service on MediaTek SoC platforms. When mtu3_gadget_queue() maps a USB transfer request via DMA before checking whether the QMU GPD ring has capacity, then mtu3_prepare_transfer() fails and returns -EAGAIN, the error path bypasses the mtu3_req_complete() cleanup helper, permanently stranding the DMA mapping. Repeated exhaustion of the IOMMU aperture can destabilize or crash the kernel. No public exploit code exists and EPSS at 0.22% (13th percentile) confirms negligible exploitation interest; the impact is a local availability-only issue confined to MediaTek USB gadget hardware.
Availability loss in the Linux kernel's USB serial digi_acceleport driver allows a local user to permanently halt data reception on a Digi AccelePort serial port until the USB device is physically reconnected or the driver is rebound. The bug arises from a missing URB resubmission path: when a port is closed while in a throttled state, the read URB is never re-queued, leaving the port silently deaf to incoming data upon next open. No public exploit has been identified and EPSS sits at 0.22% (13th percentile), reflecting the niche hardware dependency and local-only attack surface.
Hard lockup in the Linux kernel's digi_acceleport USB serial driver causes full system freeze when a Digi AccelePort device is physically disconnected during active I/O. The driver's OOB (out-of-band) command path loops indefinitely with interrupts disabled upon persistent URB submission failure, resulting in a kernel hard lockup that effectively crashes the host. Systems with this specific USB-to-serial adapter are affected across a wide range of stable kernel branches from 2.6.12 onward; patches have been released across all active stable series.
Memory leak in the Linux kernel USB ULPI (Universal Serial Bus Link Interface) subsystem allows a local low-privileged user to trigger repeated ULPI device registration failures that leak kernel memory, potentially leading to system instability or denial of service. Affected kernel versions span from 4.2 through multiple pre-patch stable branches, with fixes backported to 5.10.261, 5.15.212, 6.1.178, 6.6.145, 6.12.96, 6.18.39, 7.1.4, and 7.2-rc3. No public exploit or active exploitation has been identified at time of analysis; the EPSS score of 0.22% (13th percentile) reflects minimal probability of weaponization.
NULL pointer dereference in the Linux kernel's usbip virtual UDC (vudc) subsystem crashes the kernel when FunctionFS AIO cancellations are processed. The `vep_alloc_request()` function never initialized `vrequest->udc`, so any AIO cancellation arriving at `vep_dequeue()` - such as one triggered by a host-side `usbip attach` - dereferences a null pointer and produces a kernel oops, denying service to the guest system. A local attacker with low privileges on a system configured for USB/IP with FunctionFS AIO endpoints can reliably reproduce this crash. No public exploit code exists, and EPSS at 0.22% (13th percentile) confirms the path has not seen meaningful exploitation despite the bug reportedly existing for approximately ten years since Linux 4.7.
Kernel memory leak in the Linux USB gadget FunctionFS (f_fs) subsystem allows a local low-privileged user to slowly exhaust kernel memory by repeatedly triggering DMA buffer transfers via the ffs_dmabuf_transfer() code path. Affected kernels span from the commit introducing DMA buffer sharing (7b07a2a7ca02) through versions prior to 6.12.96, 7.1.4, and 7.2-rc3. The root cause is a missing dma_fence_put() call after dma_resv_add_fence(), leaving the initial kref permanently outstanding and causing ffs_dma_fence objects to accumulate without ever being freed. No public exploit has been identified at time of analysis.
The USB gadget FunctionFS (f_fs) subsystem in the Linux kernel fails to initialize the endpoint direction field (epfile->in) until USB host connection, leaving it zero when userspace performs early DMA ioctl calls such as FUNCTIONFS_DMABUF_ATTACH before a host connects. This incorrect DMA direction causes kernel-mode availability loss - crashes or hangs - on systems actively using FunctionFS-based USB gadgets with DMABUF. No public exploit has been identified, and EPSS probability stands at 0.21% (11th percentile), reflecting very low opportunistic exploitation risk.
The block device layer in Linux Kernel hangs indefinitely during NVMe surprise removal when `bdev_mark_dead()` is called with `surprise=true`, because a missing guard causes `sync_blockdev()` to be invoked on an already-dead device, permanently blocking `folio_wait_writeback()`. Triggered via `nvme_reset_work()` I/O-queues-lost path, the bug wedges the NVMe reset worker and every task waiting on it, requiring a system reboot to recover. Discovered by FuzzNvme (syzkaller + FEMU fuzzing framework); no public exploit identified at time of analysis.