Overview of incoming advisories.
2018 results found
Node.js released security updates on October 21 addressing 10 vulnerabilities across its three supported release lines: 22.x, 24.x, and 26.x. The most critical flaw resides in the Brotli decompression code bundled with Node.js. This vulnerability could allow remote code execution, enabling attackers to run arbitrary code on affected systems. All currently supported Node.js versions are impacted. Fixed versions have been released for each supported release line. The severity of the RCE potential makes this a high-priority patch for organizations running Node.js in production environments.
View original advisory →Security firm FoxIO disclosed critical vulnerabilities in QPACK, the header compression mechanism used in HTTP/3, that allow unauthenticated remote attackers to crash web servers with a single connection. The attack exploits a memory amplification flaw where a small batch of compressed HTTP headers expands dramatically in memory, achieving up to 4,000-fold amplification. This can exhaust server memory on default configurations, effectively enabling denial-of-service attacks. Working proof-of-concept code has been released publicly. The vulnerability affects a wide range of HTTP/3 server implementations. No authentication or prior access is required to exploit the flaw. The disclosure highlights systemic risks in the QUIC/HTTP3 protocol stack's header compression design.
View original advisory →Amazon's Threat Intelligence team detected an exploitation attempt against PHP2Shell vulnerabilities on October 17, 2026, four days before they were publicly disclosed. The attempt was observed via Amazon's MadPot honeypot service. The exploit originated from an IP address linked to a commercial anonymization network. The vulnerabilities have since come under broad exploitation in the wild following patch release. The discovery highlights the existence of pre-disclosure exploitation activity, suggesting threat actors may have had prior knowledge of the flaws. CJ Moses from Amazon commented on the findings.
View original advisory →Microsoft disclosed that China state-aligned threat actor Storm-0940 was responsible for a supply chain compromise targeting the ConnectWise Cloud Backup application. The attackers leveraged ARTEX, a Go-based agentic AI penetration testing framework, to accelerate scanning operations during the attack. ARTEX was previously made closed-source following its abuse against South Korean financial institutions. The incident represents a notable use of AI-powered offensive tooling by a nation-state actor. This supply chain attack highlights the growing sophistication of Chinese state-sponsored cyber operations and their adoption of advanced AI-driven frameworks to enhance attack efficiency.
View original advisory →Hackers compromised the infrastructure of Onamae Desktop Cloud, a major VPS service provider in Japan. The attackers installed malware on up to 26,739 customers' servers, representing a significant supply-chain-style infrastructure breach. The incident affected a large number of virtual private servers managed by the provider. Onamae is a well-known domain and hosting services company in Japan. The scale of the attack — nearly 27,000 servers — makes this a high-impact event for Japanese businesses and individuals relying on the platform. The nature of the malware installed has not been fully detailed in the article. This type of cloud infrastructure compromise can lead to data theft, further lateral movement, and deployment of additional payloads on customer environments.
View original advisory →North Korean state-sponsored hacking group Lazarus is employing post-quantum cryptographic mechanisms to secure the delivery of malicious payloads in new operations. The campaign combines both old and new malware families, demonstrating the group's continued evolution and sophistication. The use of post-quantum cryptography marks a notable advancement in their operational security, making interception and analysis of communications more difficult. This represents a proactive adaptation to future-proof their attack infrastructure against potential quantum computing decryption capabilities. The activity highlights Lazarus Group's ongoing threat to global cybersecurity and their investment in cutting-edge evasion techniques.
View original advisory →The PHP Foundation published a postmortem accusing Chainguard of breaking a coordinated security embargo by releasing patched PHP container images over six hours before the scheduled disclosure of two critical RCE vulnerabilities (CVE-2026-109618 and CVE-2026-109619). Chainguard CEO Dan Lorenc defended the early release citing active exploitation reports from multiple security vendors and the ease of independently discovering the vulnerability. The disclosure process was further disrupted by leaked CVE identifiers, an accidentally published exploit demonstration video by Calif, and potentially leaked mitigation guidance. The underlying memory management vulnerability had been known since a 2024 Quarkslab audit but stalled due to complex trade-offs and limited developer resources. Researchers from Calif, OpenAI, and Wiz contributed to identifying attack vectors and RCE proofs of concept. PHP's security team is now moving to majority-vote decisions to prevent fixes from stalling, and the Foundation is calling for increased funding from companies dependent on PHP. The incident raised community concerns about whether other projects would trust Chainguard with embargoed patches in the future.
View original advisory →CISA has issued a warning regarding active exploitation of four vulnerabilities in Plex Media Server, collectively referred to as 'Plex2Shell' attacks. Attackers are chaining these flaws together to breach servers running Plex Media Server. The advisory highlights the risk of remote code execution or unauthorized access through the chained exploitation of these vulnerabilities. CISA is urging organizations and users running Plex Media Server to apply patches immediately. The campaign demonstrates the increasing trend of attackers combining multiple lower-severity vulnerabilities to achieve high-impact outcomes. This advisory underscores the importance of timely patching for media server software exposed to the internet.
View original advisory →The popular LiteLLM Python package was compromised in a software supply chain attack affecting version 1.107.0. The malicious package was distributed via PyPI wheels and Docker images, delivering both a credential stealer and a backdoor. The compromise was discovered and reported by multiple security firms including SafeDep, Socket, StepSecurity, and Upwind. PyPI has since quarantined the affected package to prevent further installations. This incident highlights ongoing risks in open-source package ecosystems, particularly for AI/LLM tooling. Organizations using LiteLLM 1.107.0 should consider their environments potentially compromised and rotate credentials.
View original advisory →CVE-2026-109618 and CVE-2026-109619 are two critical vulnerabilities affecting PHP, a widely used scripting language powering web servers and content management systems globally. The vulnerabilities were reported by Palo Alto Networks Unit 42 as a threat brief. Due to PHP's widespread deployment, these flaws carry significant risk for web infrastructure worldwide. The article highlights the critical severity rating of both CVEs. No additional technical details such as exploit mechanisms or patch information are provided in the available content. Organizations running PHP-based web servers or CMS platforms are likely at risk. Immediate attention and patching are implied as necessary responses.
View original advisory →PHP released emergency patches on October 21, 2026, for two critical memory-safety RCE vulnerabilities tracked as CVE-2026-109618 and CVE-2026-109619, collectively known as PHP2Shell, both rated 9.2 on CVSS v4.0. The flaws affect all PHP installations serving web requests including PHP-FPM, Apache mod_php, php-cgi, FrankenPHP, and LiteSpeed, with exploitation possible before any application code executes. NSA and CERT-EU urged organizations unable to patch immediately to take affected systems offline or isolate them from public networks. Fixed versions are PHP 8.5.12, 8.4.27, 8.3.36, and 8.2.35, with minimal backports also published for PHP 7.4, 8.0, and 8.1. Cloudflare reported active attack traffic with attackers already attempting to evade WAF protections. Approximately 48.5% of PHP websites run end-of-life branches, making the patch rollout particularly challenging. Wiz researchers noted that AI models helped accelerate the development of working exploits across applications in hours. Remote memory disclosure is also possible, potentially exposing passwords, tokens, and API keys without first achieving code execution.
View original advisory →The litellm 1.107.0 release on PyPI was compromised, with backdoored wheel packages and Docker images shipping a malicious .pth file acting as a credential stealer. The malware also includes an hourly remote code loader that periodically fetches and executes additional payloads. Stolen credentials and data are exfiltrated to the domain litellm.dev. The article provides indicators of compromise (IOCs) and remediation guidance. This represents a serious supply chain attack targeting developers and organizations using the litellm Python package. The attack vector leverages Python's .pth file mechanism to achieve persistence and code execution on affected systems.
View original advisory →The PHP security team released emergency patches addressing two critical remote code execution vulnerabilities identified as CVE-2026-109618 and CVE-2026-109619, collectively tracked as PHP2Shell. These flaws allow attackers to execute arbitrary code remotely on affected PHP installations. The severity of the vulnerabilities prompted an out-of-cycle emergency release rather than waiting for a scheduled update. PHP is a widely deployed server-side scripting language powering a significant portion of the web, making these vulnerabilities high-impact. Administrators and developers are urged to apply the patches immediately to mitigate exploitation risk. The collective branding as PHP2Shell suggests the flaws may involve mechanisms that allow shell command execution through PHP. No further technical details about exploitation status were provided in the article excerpt.
View original advisory →PHP has patched two critical vulnerabilities that allow unauthenticated remote code execution on all PHP installations that process web requests. CVE-2026-109618 involves memory corruption during Out-of-Memory handling, causing internal request data to be only partially initialized and treated as valid, enabling RCE. CVE-2026-109619 involves stale internal global state persisting after memory is freed, allowing subsequent requests processed by the same worker to use outdated state before reinitialization, also leading to RCE. Both vulnerabilities are exploitable without authentication via standard web requests. All PHP installations handling web traffic are affected. The PHP team and NCSC urgently advise applying the security updates immediately.
View original advisory →Unit 42 investigated a threat actor who bypassed AI safety controls to leverage AI tools in planning and executing multiple types of cybercrime. The actor used AI assistance to coordinate credential theft operations, plan VoIP fraud schemes, and execute database exfiltration attacks. The investigation highlights the emerging risk of AI being weaponized by malicious actors to enhance their attack planning and execution capabilities. This case demonstrates how threat actors are actively finding ways to circumvent AI safety guardrails to obtain actionable cybercrime guidance. The report underscores the dual-use nature of AI technology and the need for stronger safety controls in AI platforms to prevent misuse by cybercriminals.
View original advisory →ClickFix attacks are experiencing a significant surge in activity, with cybercriminals leveraging fake AI tool installers and malicious search results to deceive users into executing malware. Attackers are impersonating trusted chatbot platforms and AI applications to make their lures more convincing and harder to detect. Push Security has analyzed how these attacks successfully evade traditional defense mechanisms. The technique relies on social engineering, tricking users into manually running malicious commands or scripts. AI-themed lures have proven particularly effective due to widespread public interest in AI tools. The campaign highlights a growing trend of threat actors exploiting emerging technology trends to increase attack success rates.
View original advisory →PHP has released emergency security updates to address two critical remote code execution vulnerabilities collectively dubbed 'PHP2Shell'. The flaws reside in the PHP runtime itself, meaning any site running PHP is potentially vulnerable regardless of the application code. Affected versions span PHP 8.2, 8.3, 8.4, and 8.5 branches, with fixes available in PHP 8.2.35, 8.3.36, 8.4.27, and 8.5.12. The vulnerabilities allow unauthenticated attackers to execute arbitrary code on web servers. Active exploit attempts have already been observed in the wild, making patching urgent. The PHP team has stated that only a full update to the patched versions provides complete protection against these flaws.
View original advisory →Researchers at Wiz, leveraging AI-assisted analysis, discovered critical remote code execution (RCE) vulnerabilities in the core PHP engine, dubbed PHP2Shell. These flaws potentially affect approximately 70% of the Internet, given PHP's widespread deployment. The vulnerabilities enable attackers to execute arbitrary code remotely on affected servers. Additionally, a Heartbleed-style memory disclosure vulnerability was identified, which could expose sensitive data from server memory. The severity and breadth of affected systems make this a critical issue comparable to historical vulnerabilities like Heartbleed. Organizations running PHP-based web infrastructure are urged to apply patches immediately. The discovery highlights the growing role of AI in identifying complex, systemic vulnerabilities in foundational web technologies.
View original advisory →A malicious version (1.107.0) of the LiteLLM package was uploaded to PyPI and Docker Hub on October 21. The compromised package contains a credential stealer targeting browser-stored credentials, cloud secrets, and AI-related secrets. Additionally, the malicious version includes a dormant command-and-control (C2) capability, suggesting potential for future remote exploitation. This is a supply chain attack targeting AI/ML developers and organizations using LiteLLM for managing large language model integrations. Users who installed version 1.107.0 should consider themselves compromised and rotate all secrets immediately. The attack highlights ongoing risks in open-source AI tooling ecosystems and package registries.
View original advisory →LiteLLM version 1.107.0 on PyPI and Docker was compromised in a supply chain attack, delivering a credential stealer via Python startup hooks (cook.pth, cook.py, sitecustomize.py). The malware collects cloud credentials (AWS, Azure, GCP), AI provider keys, SSH/GPG keys, browser data, and developer secrets. A lookalike GitHub repository 'Berri-AI/litellm' (vs legitimate 'BerriAI/litellm') was recorded in container build provenance, indicating the release workflow was hijacked. Seven wheel files across macOS, Linux, and Windows were affected, as well as multiple Docker/GHCR container images including those tagged 'latest'. Stolen data is encrypted with AES-256-CBC and RSA-4096 and exfiltrated to hxxp://www[.]litellm[.]dev/healthcheck. A remote Python code loader is controlled via a GitHub file at letters9813/pythonclient-test. Socket's AI scanner detected the malicious package seven minutes after publication. Users are advised to immediately stop using 1.107.0, revoke exposed credentials, and roll back to 1.106.2.
View original advisory →