Overview of incoming advisories.
2018 results found
CISA has added CVE-2026-21962, a maximum-severity vulnerability (CVSS 10.0) affecting Oracle HTTP Server and Oracle WebLogic Server, to its Known Exploited Vulnerabilities catalog. The flaw allows unauthenticated attackers with network access via HTTP to access critical data. Active exploitation has been confirmed in the wild. The vulnerability requires no authentication, making it especially dangerous for internet-facing deployments. Organizations using Oracle WebLogic or Oracle HTTP Server are urged to apply patches immediately per CISA's guidance.
View original advisory →CVE-2026-60004 is a code injection vulnerability in Gitea, a self-hosted Git service. An attacker with repository write access can exploit the diffpatch API endpoint by sending a malicious patch that plants an executable Git hook. This allows the attacker to execute arbitrary shell commands as the Gitea service account, potentially leading to full server compromise. The vulnerability is listed in CISA's Known Exploited Vulnerabilities catalog under BOD 26-04, which mandates prioritizing security updates based on risk. Federal agencies are required to remediate this vulnerability per BOD 26-04 directives. Forensics triage requirements are also outlined in CISA's implementation guidance. The issue is tracked at NVD and has a dedicated GitHub security advisory.
View original advisory →GitPython versions before 3.1.59 contain a vulnerability where the --separate-git-dir option is omitted from the unsafe_git_clone_options list. This oversight allows attackers to pass a separate_git_dir parameter to Repo.clone_from() or Repo.clone() methods, redirecting repository metadata to an attacker-controlled filesystem path. The flaw enables arbitrary directory creation outside the intended clone destination, and could potentially lead to hook execution if the attacker can place malicious git hooks in the redirected directory. The vulnerability is a path traversal-class issue affecting Python projects that use GitPython for git operations. A fix was introduced in GitPython version 3.1.59, and users are advised to upgrade immediately. The issue has been documented in the official GitHub security advisory and tracked by VulnCheck.
View original advisory →CVE-2026-18798 is a double free vulnerability (CWE-415) in the OpenSSL QUIC stack affecting QUIC server implementations. The flaw occurs in port_default_packet_handler() where a QRX (QUIC record layer RX) object is freed twice when port_bind_channel() fails during initial packet processing. This leads to heap corruption and typically results in termination of the QUIC server process, causing Denial of Service. The vulnerability can be triggered with relatively low effort by sending a malformed INITIAL packet with a DCID (destination connection ID) shorter than 8 bytes. While heap corruption is confirmed, remote code execution is considered highly improbable based on current analysis. The FIPS module is not affected as the QUIC implementation resides outside the OpenSSL FIPS module boundary. Patches have been issued via multiple commits to the OpenSSL repository.
View original advisory →GitPython versions before 3.1.59 contain a vulnerability where the library fails to disable merge_includes when parsing .gitmodules files. This allows attackers to craft malicious .gitmodules files containing [include] directives that point to arbitrary local files. When repo.submodules is accessed, GitConfigParser raises a MissingSectionHeaderError that embeds the first line of the targeted file verbatim in the exception message, causing unintended local file content disclosure. The vulnerability can be exploited to expose sensitive files on the host system. A fix is available by upgrading to GitPython 3.1.59 or later. The issue has been documented in both the GitHub Security Advisory and VulnCheck advisories. It is classified as a high-severity information disclosure vulnerability affecting applications that use GitPython to process potentially untrusted repositories.
View original advisory →CVE-2026-55532 affects PraisonAI, a multi-agent teams system, in versions prior to 4.6.58. The vulnerability exists in the MCP HTTP Stream _validate_origin function, which uses a startswith() check to validate request origins against an allowlist. An attacker can craft a malicious origin such as localhost.attacker.com that satisfies the localhost allowlist check due to the flawed string prefix matching. This allows cross-origin requests to be sent without CORS preflight by using Content-Type: text/plain, enabling unauthenticated access to the tools/call API endpoint without an API key. Successful exploitation can lead to unauthorized file writes that persist malicious agent instructions, potentially compromising the integrity of AI agent behavior. The vulnerability has been patched in PraisonAI version 4.6.58.
View original advisory →A heap out-of-bounds read vulnerability exists in OpenEXRUtil versions 3.3.0 through 3.3.12 and 3.4.0 through 3.4.12. The flaw is in the TypedDeepImageChannel<T>::row() API, which can return an out-of-bounds pointer when a deep image has a non-zero dataWindow origin. The root cause is a conflict between two coordinate models in ImfDeepImageChannel: at(x, y) uses absolute coordinates while row(r) is documented as 0-based logical access. For non-zero dataWindow.min values, row(0) points outside the _sampleListPointers allocation, causing a heap out-of-bounds read and potential crash. Under a controlled heap layout, this could lead to information disclosure. The vulnerability affects the EXR image format library widely used in the motion picture industry. Fixes are available in versions 3.3.13 and 3.4.13.
View original advisory →CVE-2026-55585 affects the QWED open-source AI verification package prior to version 5.1.2. The vulnerability stems from passing caller-controlled math expressions directly to SymPy's parse_expr() function without restricted namespaces, enabling arbitrary Python code execution via eval(). Two attack vectors exist: POST /verify/math and POST /verify/batch, both reachable with any valid tenant API key. The default-enabled signup endpoint allows anyone to self-register and obtain an API key, making the vulnerability unauthenticated in practice. Exploitation can lead to file read/write, OS command execution, service termination, and cross-tenant compromise in shared deployments. The issue is fixed in QWED version 5.1.2.
View original advisory →Grav CMS versions before 2.0.16 contain a symlink following vulnerability in the Scheduler Job::createLockFile() function. Local attackers can exploit this by pre-creating symlinks at predictable lock file paths within the world-writable temp directory. When a scheduled job runs, it follows the symlink and overwrites the target file with a job ID string. The attacker can point the symlink to any file writable by the web server process, enabling arbitrary file overwrite. This vulnerability requires local access but can have significant impact on system integrity. The issue has been patched in Grav CMS version 2.0.16. Multiple advisories have been published including on GitHub Security Advisories and VulnCheck.
View original advisory →Crater Invoice through version 6.0.6 contains a critical path traversal vulnerability in its self-update API. Authenticated company owners can exploit this flaw by supplying crafted ZIP archives containing ../ sequences to the unzip endpoint. The vulnerability stems from unsanitized ZIP entry names being passed directly to PHP's ZipArchive::extractTo() function. This allows attackers to write arbitrary files outside the intended extraction directory, including PHP files into web-accessible public directories. Successful exploitation results in remote code execution (RCE) on the server. The vulnerability requires authentication as a company owner, slightly limiting the attack surface but still posing a severe risk in multi-tenant or compromised credential scenarios. Public proof-of-concept and advisories are available, increasing exploitation risk.
View original advisory →A heap out-of-bounds write vulnerability exists in OpenEXR, the reference implementation for the EXR image format used widely in the motion picture industry. The flaw affects versions before 3.2.11, 3.3.0 through 3.3.12, and 3.4.0 through 3.4.13 on 32-bit/ILP32 builds. A specially crafted tiled EXR file with a small 40x40 dataWindow but an oversized 65537x65537 tile size can trigger the vulnerability through the TiledRgbaInputFile RGBA API. On ILP32 systems, the Array2D tile-conversion buffer size calculation overflows, causing allocation of a much smaller heap buffer than required. Tile decode operations then write past the allocated buffer boundary, resulting in heap memory corruption. The vulnerability has been patched in versions 3.2.11, 3.3.13, and 3.4.14. Exploitation requires an attacker to supply a maliciously crafted EXR file to a vulnerable application running on a 32-bit/ILP32 platform.
View original advisory →CVE-2026-79664 affects Ech0 versions before 4.7.3, where access tokens created with the never-expire option cannot be properly revoked. Three separate revocation mechanisms all fail: the logout function panics on a nil ExpiresAt field, RevokeToken skips revocation when remainTTL is zero, and the admin delete function does not blacklist the JWT ID (JTI). As a result, stolen tokens remain cryptographically valid indefinitely unless the JWT secret is rotated. This allows attackers who obtain a never-expire token to maintain persistent authenticated access to the system. The vulnerability is classified as high severity due to the potential for perpetual unauthorized access. A fix is available in Ech0 version 4.7.3 and later.
View original advisory →CVE-2026-24262 is a vulnerability discovered in the system firmware of NVIDIA DGX Spark. A privileged attacker can exploit an out-of-bounds write flaw in the firmware. Successful exploitation may result in arbitrary code execution, privilege escalation, denial of service, information disclosure, and data tampering. The vulnerability requires privileged access to exploit, somewhat limiting the attack surface. NVIDIA has published a security advisory through their product-security GitHub repository. The vulnerability is tracked by both NVD and CVE.org. Given the range of potential impacts including code execution and privilege escalation, this is considered a high-severity issue. The DGX Spark is an AI computing platform, making this vulnerability particularly relevant to high-performance AI infrastructure environments.
View original advisory →The getgrav/grav-plugin-login Composer plugin before version 3.9.1 contains a timing side-channel vulnerability in its password reset and account activation token comparison logic. The affected code uses a non-constant-time === string comparison instead of the secure hash_equals() function in classes/Controller.php and login.php. Additionally, the token submission endpoint (taskReset) lacks rate limiting, compounding the risk. An attacker with knowledge of a valid username could theoretically send repeated token guesses and use timing differences to recover a valid reset or activation token. The vendor acknowledges the flaw but rates practical exploitability as low, and no end-to-end network exploit has been publicly demonstrated. The fix is available in version 3.9.1 of the plugin.
View original advisory →PraisonAI, a multi-agent teams system, contains a critical vulnerability in its Jobs API where the create_app function mounts the /api/v1/runs endpoint without any authentication. Prior to version 4.6.51, any reachable caller could exploit this to submit jobs, read results, cancel runs, or delete jobs using operator-level credentials. This effectively grants unauthorized users full control over job operations without requiring any credentials. The vulnerability poses significant risk as it allows unauthenticated access to sensitive job management functionality. The fix introduces PRAISONAI_JOBS_API_KEY middleware that enforces authentication via Authorization or X-API-Key headers. The issue has been patched in version 4.6.58 of PraisonAI.
View original advisory →NVIDIA NemoClaw for Linux contains a critical vulnerability in its status and logs plugin commands that allows OS command injection. An attacker exploiting this flaw could achieve arbitrary code execution on the affected system. The vulnerability also enables data tampering, information disclosure, and denial of service attacks. The issue resides specifically in plugin command handling within the NemoClaw application. NVIDIA has published a security advisory referencing this CVE. The vulnerability has been assigned a high criticality rating given the potential for full system compromise.
View original advisory →CVE-2026-55528 affects PraisonAI, a multi-agent teams system, in versions prior to praisonaiagents 1.6.58. The vulnerability exists because AgentServer exposes ServerConfig.auth_token but the _create_app method fails to enforce authentication checks on any route. This allows remote unauthenticated callers to subscribe, publish, and perform other sensitive actions without providing a valid bearer token or X-Auth-Token header, even when authentication is explicitly configured. The flaw represents a complete authentication bypass in the AgentServer component. The issue has been resolved in praisonaiagents version 1.6.58. Users are strongly advised to upgrade immediately to mitigate the risk of unauthorized access to their multi-agent systems.
View original advisory →CVE-2026-16231 affects hbs, an Express view engine wrapping Handlebars, versions 2.1.0 through 4.2.1. The vulnerability exists in the registerAsyncHelper API, which bypasses Handlebars' automatic HTML escaping mechanism. During the first render pass, async helpers return an opaque placeholder that gets escaped, but the actual return value is substituted without escaping after rendering. This flaw affects cached, uncached, and layout render paths. Attackers can inject arbitrary HTML and JavaScript into server-rendered pages if attacker-controlled data is passed into async helper callbacks. The vulnerability enables both stored and reflected cross-site scripting (XSS) attacks. The issue is fixed in version 4.3.0, which properly HTML-escapes async helper output. Applications requiring raw HTML output can use hbs.SafeString to explicitly opt in. Users are strongly advised to upgrade to version 4.3.0 immediately.
View original advisory →NLTK versions before 3.10.3 contain a critical remote code execution vulnerability in their allowlisted pickle loaders. The flaw exists because the loaders trust entire module namespaces rather than restricting access to specific safe callables. Attackers can craft malicious pickle payloads that invoke dangerous functions such as ReppTokenizer._execute and numpy.f2py.crackfortran.myeval via the pickle REDUCE opcode. Exploitation occurs during the loading of model or tokenizer artifacts, making supply chain and model-sharing scenarios particularly risky. The vulnerability allows arbitrary command execution on affected systems. Users are urged to upgrade to NLTK 3.10.3 or later to remediate the issue. Details are available via NVD, the NLTK GitHub security advisory, and VulnCheck.
View original advisory →The Total Donations plugin for WordPress contains a critical privilege escalation vulnerability affecting all versions up to and including 2.0.5. Unauthenticated attackers can exploit this flaw to elevate their privileges to administrator level without any prior authentication. The vulnerability has been assigned CVE-2026-78570 and is documented by both NVD/NIST and security researchers at Patchstack and Wordfence. No patch details are explicitly mentioned, but the affected version ceiling is 2.0.5. The ability for unauthenticated users to gain admin access makes this a high-severity issue for any WordPress site running the affected plugin. Site administrators should immediately assess their exposure and apply any available updates or mitigations.
View original advisory →