Overview of incoming advisories.
2018 results found
A vulnerability in OpenEXR, the reference implementation for the EXR image format used in the motion picture industry, allows crafted EXR files with a nonzero dataWindow.min to cause TypedFlatImageChannel::row() to return an invalid heap pointer. This results in out-of-bounds or use-after-free writes when an application writes rows through FlatHalfChannel::row(). Affected versions include those before 3.2.11, 3.3.0 through 3.3.12, and 3.4.0 through 3.4.13. Tools, converters, render pipeline components, and image-processing services that accept untrusted EXR files are at risk. The vulnerability can be triggered by processing maliciously crafted EXR files from untrusted sources. Fixes have been released in versions 3.2.11, 3.3.13, and 3.4.14. Users are advised to upgrade to the patched versions immediately to mitigate the risk of memory corruption attacks.
View original advisory →A vulnerability has been identified in the system firmware of NVIDIA DGX Spark, tracked as CVE-2026-47626. The flaw allows a privileged attacker to trigger an out-of-bounds write condition within the firmware. Successful exploitation can result in a range of severe impacts including arbitrary code execution, escalation of privileges, 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 via their product-security GitHub repository. The vulnerability is catalogued by both NVD (National Vulnerability Database) and the CVE Program. Given the potential for code execution and privilege escalation at the firmware level, the severity is considered high. Organizations using NVIDIA DGX Spark hardware should monitor for firmware patches and apply mitigations as soon as available.
View original advisory →Grav Flex Objects plugin versions 1.4.0 through 1.4.7 contain an authorization bypass vulnerability in the flex-objects shortcode handler. Users with page-edit access can render any registered Flex collection without proper permission checks. Attackers can embed the shortcode in published pages to expose sensitive directory contents, including user account information. The vulnerability bypasses the authorize ACL enforced in the admin panel. This represents a privilege escalation risk where lower-privileged users can access data beyond their authorization scope. The flaw is specifically tied to the shortcode rendering path, which lacks the same access controls applied elsewhere. Organizations running affected versions of the plugin on Grav CMS installations should prioritize patching. References include the NVD entry, a GitHub security advisory, and a VulnCheck advisory.
View original advisory →CVE-2026-56705 affects Adminer versions before 5.4.3, which fails to sanitize the server field before constructing a PDO DSN string. This allows unauthenticated attackers to inject ODBC parameters using semicolons. By injecting TraceFile and TraceOn parameters, attackers can write arbitrary PHP code to the web root. When the resulting trace file is subsequently accessed via the web server, remote code execution is achieved. The vulnerability requires no authentication, making it particularly dangerous for publicly exposed Adminer instances. The fix is available in Adminer 5.4.3. The issue has been documented by both the GitHub security advisory and VulnCheck.
View original advisory →The Total Donations plugin for WordPress contains a SQL Injection vulnerability affecting all versions up to and including 2.0.5. The flaw stems from insufficient escaping of user-supplied parameters and inadequate preparation of existing SQL queries. Unauthenticated attackers can exploit this vulnerability by appending additional SQL queries to extract sensitive information from the database. No authentication is required to exploit this vulnerability, making it particularly dangerous for any site running the affected plugin versions. Users are advised to update or remove the plugin to mitigate the risk of data exfiltration.
View original advisory →CVE-2026-24169 affects NVIDIA UFM Enterprise, specifically its plugin management API. An authenticated user with low privileges can exploit this vulnerability by sending a specially crafted API request to inject arbitrary code. Successful exploitation can result in remote code execution, privilege escalation, and information disclosure. The vulnerability highlights a critical risk in NVIDIA's network fabric management platform, which is commonly used in high-performance computing and data center environments. The low privilege requirement for exploitation increases the likelihood of abuse by insider threats or compromised accounts. NVIDIA has published a security advisory via their product-security GitHub repository. The vulnerability has been assigned a high criticality rating.
View original advisory →Winter CMS versions prior to 1.2.13 contain an incomplete fix for a Twig sandbox escape vulnerability in the System\Twig\SecurityPolicy component. Authenticated backend users with template-editing permissions can bypass sandbox restrictions through method forwarding via Eloquent models and query builders. Exploitable methods include saveQuietly(), deleteQuietly(), increment(), decrement(), and newQuery(). Successful exploitation allows attackers to read and modify arbitrary database records, execute arbitrary SQL queries, and achieve remote code execution by injecting PHP code into template sections. The vulnerability represents an incomplete patch for a previously identified sandbox escape issue. A fix is available in Winter CMS version 1.2.13 and later.
View original advisory →CVE-2026-55526 affects PraisonAI's multi-agent system (praisonaiagents) prior to version 1.6.58. The vulnerability exists in spider_tools._host_is_blocked(), which fails to resolve hostnames before scrape_page fetches them. An attacker can craft a hostname like 127.0.0.1.nip.io that passes the blocklist validation but resolves to a loopback address, enabling Server-Side Request Forgery (SSRF) to access internal HTTP services. This is a classic DNS rebinding/SSRF bypass technique. The fix was implemented in version 1.6.58 by using socket.getaddrinfo to resolve hostnames prior to validation and failing closed on DNS errors. Users should upgrade to praisonaiagents 1.6.58 or later immediately to remediate the issue.
View original advisory →CVE-2026-55538 affects PraisonAI, a multi-agent teams system, prior to version 4.6.51. The vulnerability exists because the _create_agents_app() function fails to enforce authentication on POST /agents and POST /agents/{agent_name} endpoints, despite the system parsing an API key from config. As a result, requests with missing or incorrect bearer tokens or X-API-Key headers are still processed and reach agent execution. This constitutes a broken authentication vulnerability that could allow unauthorized actors to interact with and execute agents without valid credentials. The issue has been patched in version 4.6.58 of PraisonAI. Users are advised to upgrade immediately to mitigate the risk of unauthorized agent access and potential misuse of the multi-agent system.
View original advisory →CVE-2026-55640 affects the Nextcloud MCP Server, a production-ready server connecting AI assistants to Nextcloud instances. Prior to version 0.117.2, the POST /webhooks/nextcloud endpoint had no authentication by default because the WEBHOOK_SECRET environment variable defaulted to None and startup validation did not enforce its presence. When WEBHOOK_SECRET is unset, the handle_nextcloud_webhook() function accepts unauthenticated requests from any network attacker. The attacker-controlled payload field payload['user']['uid'] is parsed without an authenticated-session cross-check before being used in Qdrant vector database operations. This allows a remote unauthenticated attacker to delete or trigger re-indexing of vector embeddings for any user and to destroy the semantic search index by sending forged deletion events. The vulnerability is fixed in version 0.117.2.
View original advisory →CVE-2026-55581 affects mcp-shell, an MCP server designed for secure and auditable shell command execution. Prior to version 0.6.0, the default Docker security configuration included /bin/bash in the allowed_executables list. The security validation logic in security.go only checked the first token of a command, and the checkBlockedPatternsAndCommands function failed to reject the -c flag used in shell command mode. An attacker could exploit this by supplying '/bin/bash -c <arbitrary-command>' as the command argument to the shell_exec MCP tool, bypassing the intended allowlist. The malicious command would pass validation and be executed by executor.go via exec.CommandContext as the mcpuser. This effectively allowed arbitrary command execution outside the defined security policy. The vulnerability has been patched in version 0.6.0.
View original advisory →Nokogiri versions before 1.11.4 (CRuby implementation using packaged/vendored libxml2) bundle libxml2 2.9.10, which contains multiple known vulnerabilities fixed in libxml2 2.9.12. The bundled vulnerable library exposes users to CVE-2019-20388 (memory leak in xmlSchemaValidateStream), CVE-2020-24977 (global buffer over-read in xmlEncodeEntitiesInternal), CVE-2021-3517 (heap-based buffer overflow), and CVE-2021-3518 (out-of-bounds read). Exploitation occurs through processing of specially crafted XML documents. Potential impacts include denial of service, information disclosure, and memory corruption. The vulnerability is a supply chain/dependency issue stemming from vendored third-party library usage. Fixes are available via multiple commits to the Nokogiri repository and upgrading to version 1.11.4 or later is recommended.
View original advisory →A prototype pollution vulnerability (CVE-2026-78654) has been discovered in the cleverbrush framework and deep library up to version 4.4.0. The vulnerability exists in the deepExtend function within libs/deep/src/deepExtend.ts, allowing improperly controlled modification of object prototype attributes. Remote exploitation is possible and a public exploit has been disclosed. The vulnerability has been patched in version 4.4.1 with commit 810398c1308c500c3b8b6af380b5a89371389327. Users are strongly advised to upgrade the affected component immediately to mitigate risk.
View original advisory →CVE-2026-55620 affects the eml_parser Python library (versions prior to 3.0.2), which is used for parsing EML email files. The vulnerability exists in the eml_parser.routing.noparenthesis function in routing.py, where a regex-based fix-point loop processes parenthesized CFWS comments in Received headers with quadratic time complexity relative to nesting depth. An attacker can craft a small EML file containing deeply nested parentheses (e.g., 5,000 levels) to cause approximately 1.3 seconds of CPU saturation per message, with processing time quadrupling as nesting depth doubles. This ReDoS (Regular Expression Denial of Service) vulnerability can lead to worker latency, queue backpressure, and service-level outages in synchronous gateways, sandboxes, and real-time triage pipelines. The issue has been patched in eml_parser version 3.0.2. The fix is available via a GitHub commit and documented in a security advisory on GitHub.
View original advisory →CVE-2026-65081 is a high-severity vulnerability affecting NVIDIA NemoClaw for Linux. The flaw exists within the software's installation process, allowing an attacker to cause execution of untrusted code. A successful exploit could lead to multiple severe impacts including arbitrary code execution, escalation of privileges, data tampering, information disclosure, and denial of service. The vulnerability was reported via the NVD (National Vulnerability Database) and NVIDIA's product security advisory. No patch or mitigation details are explicitly provided in the article content. The vulnerability is classified as high criticality given the breadth of potential impacts. Organizations using NVIDIA NemoClaw on Linux systems should monitor NVIDIA's security advisories for updates and patches.
View original advisory →CVE-2026-65082 is a vulnerability in NVIDIA NemoClaw for Linux affecting its migration command. A local attacker can exploit this flaw to perform code injection. Successful exploitation may lead to arbitrary code execution, data tampering, information disclosure, and denial of service. The vulnerability is locally exploitable, meaning an attacker must have local access to the system. NVIDIA has published a security advisory via their product-security GitHub repository. The vulnerability is tracked on both the NVD and CVE.org databases. The severity is rated High based on the range of potential impacts. No patch or workaround details are provided in the article content itself.
View original advisory →Nokogiri versions prior to 1.13.5 contain an integer overflow vulnerability in the packaged libxml2 buffer handling functions. The flaw enables attackers to cause out-of-bounds memory writes by crafting multi-gigabyte XML files that trigger buffer overflows. Successful exploitation can lead to information disclosure, unauthorized data modification, or denial of service conditions. The vulnerability originates in the underlying libxml2 library bundled with Nokogiri, a widely used Ruby XML/HTML parsing library. A patch was issued in Nokogiri 1.13.5, and users are advised to upgrade immediately. The issue is tracked under CVE-2022-50999 and has been assigned a High criticality rating. Multiple references including a GitHub commit, a security advisory, and a VulnCheck entry document the vulnerability and its remediation.
View original advisory →CVE-2026-65097 describes a vulnerability in NVIDIA NemoClaw for Linux affecting its installation scripts. The flaw allows an attacker to trigger a download of code without integrity verification, commonly known as a 'download without integrity check' weakness. Successful exploitation could lead to arbitrary code execution, privilege escalation, information disclosure, and data tampering. The vulnerability resides specifically in the installation/setup phase of the software. NVIDIA has published a security advisory via their product-security GitHub repository. The vulnerability is rated high severity given the breadth of potential impact. Linux systems running NVIDIA NemoClaw are the primary affected targets. Users are advised to review NVIDIA's official advisory for patches and mitigations.
View original advisory →Nokogiri versions prior to 1.19.3 are vulnerable to Regular Expression Denial of Service (ReDoS) in the CSS selector tokenizer. The vulnerability affects string-literal and identifier tokenization within the library. Attackers can exploit this by injecting adversarial CSS selectors into methods such as Node#css, Node#at_css, and Searchable#search. This causes exponential regex backtracking, leading to denial of service conditions. The vulnerability is tracked as CVE-2026-79770 and has been patched in Nokogiri 1.19.3. Users of affected versions should upgrade immediately to mitigate the risk. The issue is documented across NVD, GitHub Security Advisories, and VulnCheck.
View original advisory →A vulnerability was discovered in the sos clean utility, part of the sos package used for system diagnostics. The flaw allows a local attacker to perform arbitrary file creation or overwrite by crafting a malicious tar archive. The vulnerability stems from improper validation of symlink and hardlink targets during tar extraction, enabling path traversal attacks. Since the sos clean process frequently runs with root privileges, a successful exploit can write files to any location on the system. This poses a significant privilege escalation and system integrity risk. The issue has been tracked and reported via Red Hat's Bugzilla and the upstream GitHub repository. Fixes are being addressed through a pull request in the sosreport/sos GitHub project. The vulnerability affects Linux systems running the sos package, particularly Red Hat-based distributions.
View original advisory →