Overview of incoming advisories.
1553 results found
CVE-2026-55609 is a path traversal vulnerability affecting the sublinear-time-solver Rust/WebAssembly library and its consciousness-explorer component. The export_state, import_state, saveVectorToFile, and loadVectorFromFile MCP tools pass attacker-controlled filepath parameters directly to filesystem operations without sanitization or traversal restrictions. An attacker capable of invoking these MCP tools can read, write, or overwrite any file accessible to the server process. This leads to confidentiality and integrity loss and potential service disruption. The vulnerability exists in src/consciousness-explorer/mcp/server.js and src/mcp/server.ts. Fixes are available in consciousness-explorer 1.1.2 and sublinear-time-solver 1.6.0. Multiple GitHub commits and a pull request address the remediation.
View original advisory →A Server-Side Request Forgery (SSRF) vulnerability exists in PraisonAI's multi-agent system prior to version praisonaiagents 1.6.58. The web_crawl function validates only the initial URL but uses httpx.Client with follow_redirects=True, meaning redirect targets are never revalidated. An attacker who can influence a crawl target can craft a redirect from a public URL to internal loopback addresses, private network ranges, or cloud metadata services (e.g., AWS IMDSv1). This bypass works even when ALLOW_LOCAL_CRAWL is disabled, as the security check only applies to the initial URL. The internal response content is returned directly to the agent context, potentially exposing sensitive infrastructure data. The vulnerability is fixed in praisonaiagents version 1.6.58.
View original advisory →NLTK versions before 3.10.3 contain a server-side request forgery (SSRF) vulnerability in nltk.pathsec.urlopen and its callers (nltk.data.load, nltk.downloader.Downloader.index/download) when an HTTP proxy is configured. The vulnerability arises because pathsec.urlopen validates the requested hostname locally, but proxy-handler inheritance disables the safe HTTP/HTTPS handlers, causing the actual fetch to be performed by the proxy without re-validation of the destination. An attacker can supply a publicly validated URL that the proxy then forwards to an internal loopback-only service. This allows disclosure of internal HTTP resources, loading of forged downloader indexes, and installation of attacker-chosen package content. The flaw represents a significant supply chain risk for Python NLP environments using NLTK with proxy configurations. Users are advised to upgrade to NLTK 3.10.3 or later to mitigate the vulnerability.
View original advisory →OpenEXR, the reference implementation for the EXR image format widely used in the motion picture industry, contains a vulnerability in multiple versions. Affected versions include those before 3.2.11, 3.3.0 through 3.3.12, and 3.4.0 through 3.4.13. The flaw resides in the TypedDeepImageChannel::row() function, which can return an out-of-bounds pointer when processing a crafted deep EXR file with a nonzero dataWindow origin. The root cause is an API design issue where zero-based row access is combined with an absolute-coordinate-adjusted base pointer. Exploitation of this vulnerability can lead to application crashes or limited information disclosure. Fixed versions are 3.2.11, 3.3.13, and 3.4.14. Multiple commits have been published to the OpenEXR GitHub repository addressing this issue. A GitHub Security Advisory (GHSA-6662-fq6f-93mp) has also been published.
View original advisory →CVE-2026-65090 is a vulnerability in NVIDIA NemoClaw for Linux affecting the NIM management component. The flaw allows an attacker to perform OS command injection through the affected component. Successful exploitation could lead to arbitrary code execution, data tampering, information disclosure, and denial of service. The vulnerability is rated as high severity. NVIDIA has published a security advisory through their product-security GitHub repository. The issue is tracked in both the NVD and CVE databases. No additional technical details or proof-of-concept exploits are currently disclosed in the article. Users of NVIDIA NemoClaw on Linux should monitor for patches and apply mitigations as soon as they become available.
View original advisory →CVE-2026-55582 affects mcp-shell, an MCP server for running shell commands securely. Prior to version 0.6.0, the default security.yaml configuration permits /usr/bin/git, while the security.go file fails to include the '!' character in its shell metacharacter and dangerous construct checks. This oversight, combined with no per-executable argument policy, allows an attacker to pass a crafted Git -c alias argument to execute arbitrary OS commands. The exploit leverages Git's alias feature with shell escape syntax (!<command>) to bypass security controls. The default Docker deployment runs as mcpuser with Git installed and secure mode enabled, making it exploitable without additional authentication beyond MCP connectivity. The vulnerability is exploitable in default configurations, raising its risk profile significantly. The issue has been patched in version 0.6.0 of mcp-shell. Users are advised to upgrade immediately to mitigate the risk of arbitrary command execution.
View original advisory →CVE-2026-65099 is a vulnerability in NVIDIA NemoClaw for Linux affecting its command-line interface. The flaw allows an attacker to perform OS command injection through the CLI. Successful exploitation can lead to arbitrary code execution on the affected system. Additionally, the vulnerability may result in data tampering, undermining data integrity. Information disclosure is also a potential consequence, exposing sensitive data. Denial of service attacks are possible, disrupting system availability. NVIDIA has published an advisory on their product-security GitHub repository. The vulnerability is tracked on both NVD and CVE.org. No active exploitation details are mentioned in the advisory at this time.
View original advisory →NVIDIA NemoClaw for Linux contains a critical vulnerability in its inference server setup that allows remote attackers to access the inference service without authentication. The flaw can be exploited remotely without any credentials, making it particularly dangerous in exposed deployments. A successful exploit may lead to information disclosure, potentially exposing sensitive model data or inference results. The vulnerability also enables denial of service attacks, which could disrupt AI inference workloads. NVIDIA has published a security advisory on their GitHub product-security repository. The vulnerability has been assigned CVE-2026-65105 and is tracked on both NVD and CVE.org. Given the unauthenticated remote access nature, this vulnerability poses a high risk to organizations using NVIDIA NemoClaw in their AI/ML infrastructure.
View original advisory →Nokogiri versions before 1.13.2 (CRuby with packaged libraries) ship vendored libxml2 2.9.12 and libxslt 1.1.34, both affected by upstream CVEs. CVE-2021-30560 in libxslt exposes applications to denial-of-service when transforming XML with untrusted XSL stylesheets. CVE-2022-23308 in libxml2 exposes applications to denial-of-service, memory disclosure, or code execution when parsing untrusted documents with specific parse options (DTDVALID=true, NOENT=false). These vulnerabilities are inherited through Nokogiri's vendored/bundled dependencies, representing a supply chain risk. The fix is to upgrade to Nokogiri 1.13.2, which updates libxml2 to 2.9.13 and libxslt to 1.1.35. Applications using older versions of Nokogiri with CRuby and packaged libraries should upgrade immediately. Multiple GitHub commits and security advisories have been issued to address these issues.
View original advisory →Adobe Campaign Classic (ACC) is affected by a critical OS Command Injection vulnerability (CVE-2026-76195) classified under CWE as Improper Neutralization of Special Elements used in an OS Command. The vulnerability allows an attacker to execute arbitrary code in the context of the current user. No user interaction is required for exploitation, making it particularly dangerous. The scope of the vulnerability is changed, indicating impact can extend beyond the vulnerable component. Adobe has published a security advisory (APSB26-134) addressing this issue. The vulnerability is tracked on NVD with a High criticality rating. Organizations using Adobe Campaign Classic should apply patches immediately.
View original advisory →CVE-2026-78676 affects GitPython versions before 3.1.59, where the library fails to safely re-serialize multi-line git-config values during write operations. This flaw allows dormant quoted values containing embedded newlines to be corrupted into active git directives such as core.hooksPath. An attacker can craft a malicious git config file that appears benign until any unrelated GitPython config write operation triggers the injection. Once activated, the injected directives can redirect git hooks to attacker-controlled scripts, enabling arbitrary code execution. The vulnerability is particularly dangerous in CI/CD pipelines and developer environments where GitPython is commonly used. No user interaction beyond a normal config write is required to activate the payload. The fix is available in GitPython 3.1.59 and users are strongly advised to upgrade immediately.
View original advisory →CVE-2026-55553 affects the urllib HTTP client library for Node.js, which incorrectly reuses caller-supplied options including authentication headers when following redirects across different origins. The vulnerability exists in src/HttpClient.ts where the #requestInternal function recursively passes the same options object to redirected requests regardless of scheme, host, or port changes. Sensitive headers such as Authorization, Cookie, Proxy-Authorization, x-api-key, x-auth-token, and x-access-token can be leaked to attacker-controlled redirect destinations. No user interaction is required for exploitation, making this a significant credential exposure risk. The issue affects urllib versions prior to 2.44.1 and 4.9.1. Fixes have been released in versions 2.44.1 and 4.9.1 via two separate commits. The vulnerability could allow attackers to harvest credentials and potentially reuse them against the original API or related services.
View original advisory →CVE-2026-63075 is a vulnerability in OpenSSL's QUIC stack where a remote peer can trigger unbounded memory growth leading to Denial of Service. The flaw arises because OpenSSL retains metadata for ACK-only packets even though the QUIC protocol does not require peers to acknowledge such packets. A malicious peer can exploit this by repeatedly sending PING frames to force ACK-only responses while withholding acknowledgements, causing persistent memory accumulation. This attack only requires completing a valid QUIC handshake, making it accessible to remote unauthenticated attackers. The impact is especially severe under sustained traffic or with many concurrent QUIC connections. The vulnerability is classified under CWE-770 (Allocation of Resources Without Limits or Throttling). The fix involves updating watermark accounting for ACK-only packets without storing their metadata. The OpenSSL FIPS module is not affected as QUIC code is outside the FIPS boundary. Multiple commits have been issued across OpenSSL branches to remediate this issue.
View original advisory →Nokogiri versions before 1.18.3 are affected by two vulnerabilities rooted in libxml2. The first is a stack buffer overflow triggered when reporting DTD validation errors involving long QName prefixes. The second is a use-after-free vulnerability that occurs during validation against untrusted XML Schemas (XSD files). Both vulnerabilities can be exploited by providing malicious DTD content or untrusted XSD files to affected applications. Successful exploitation may result in denial of service or potential arbitrary code execution. The vulnerabilities are tracked under CVE-2025-71407 and have been patched in Nokogiri 1.18.3. Users and maintainers relying on Nokogiri for XML parsing and validation should upgrade immediately.
View original advisory →CVE-2026-68513 describes a heap buffer overflow vulnerability in PyOpenEXR, the Python wrapper for the OpenEXR image format library used in the motion picture industry. The flaw affects versions 3.3.0–3.3.12 and 3.4.0–3.4.13 and is triggered by a channel-name key collision between literal and prefixed RGB channels. When separate_channels=false, the channelNameToRGBA() function fails to properly handle crafted EXR files containing both literal channels (e.g., 'left') and prefixed channels (e.g., 'left.R', 'left.G', 'left.B'). This collision causes the wrapper to reuse an undersized NumPy array, resulting in an out-of-bounds write during pixel decoding. An attacker could exploit this by supplying a specially crafted flat scanline EXR file. The vulnerability has been patched in versions 3.3.13 and 3.4.14.
View original advisory →NLTK versions prior to 3.10.3 are vulnerable to a denial-of-service attack via XML entity expansion, commonly known as a 'Billion Laughs' attack. The library uses Python's xml.etree.ElementTree to parse XML across multiple modules, which processes entity declarations defined in document DTDs. Attackers can craft malicious XML payloads containing deeply nested entity declarations that expand exponentially in memory, turning small payloads of hundreds of bytes into megabytes of data. This memory exhaustion can cause denial of service on affected systems. The fix is available in NLTK version 3.10.3. The vulnerability has been documented in the NLTK GitHub security advisories and VulnCheck advisories. Users are strongly advised to upgrade to the patched version immediately.
View original advisory →CVE-2026-24263 affects NVIDIA DGX Spark system firmware, where a privileged attacker can trigger a NULL pointer dereference. Successful exploitation may lead to code execution, privilege escalation, denial of service, information disclosure, and data tampering. The vulnerability requires privileged access to exploit, limiting the attack surface somewhat. However, the breadth of potential impact makes this a high-severity issue. NVIDIA has published an advisory via their product-security GitHub repository. The vulnerability is tracked by both the NVD and CVE databases. Users of NVIDIA DGX Spark systems should apply firmware updates as soon as they become available.
View original advisory →CVE-2026-75496 affects Webkul QloApps, a hotel booking and management platform. The vulnerability stems from improper validation of uploaded file extensions and MIME types before files are moved to a publicly accessible directory. A remote, authenticated attacker with administrative privileges can exploit this flaw to upload executable files and achieve remote code execution on the server. The issue has been patched in commit 153ec1c. The vulnerability is classified as high severity due to its potential for full server compromise, though exploitation requires authenticated administrative access.
View original advisory →CVE-2026-34968 affects Adminer versions before 5.4.3, exposing a critical arbitrary file deletion vulnerability in SQLite mode. The flaw exists in the database-list drop action, which fails to validate file extensions before performing deletions. An authenticated attacker can exploit this by submitting arbitrary relative file paths via the db[] parameter. This allows deletion of any file writable by the PHP process running Adminer. The vulnerability requires authentication but can have severe impact on server integrity and availability. It was patched in Adminer version 5.4.3. Multiple advisories have been published including on GitHub Security Advisories and VulnCheck. Organizations using Adminer in SQLite mode should upgrade immediately to mitigate risk.
View original advisory →Nokogiri versions before 1.14.3 (CRuby implementation using packaged libxml2) bundle libxml2 v2.10.3, which contains a NULL pointer dereference vulnerability in XML Schema processing functions. The affected functions include xmlSchemaFixupComplexType and xmlSchemaCheckCOSSTDerivedOK. An attacker can supply a crafted or malformed XML schema to trigger the NULL pointer dereference, potentially causing a segfault and resulting in a denial of service condition. The vulnerability is tracked as CVE-2023-28484 in libxml2. Nokogiri 1.14.3 resolves the issue by upgrading the bundled libxml2 to version 2.10.4. Only users of the CRuby implementation with the packaged libxml2 are affected.
View original advisory →