Remote code execution vulnerability in SharePoint Server arising from deserialization of untrusted data (CVSS 8.8). CISA added to KEV following confirmed active exploitation by Storm-2603 threat actor deploying Warlock ransomware. Any authenticated attacker with Site Member permissions can trigger RCE without elevated privileges.
Apply Microsoft's security update for CVE-2026-45659 immediately. Review logs for exploitation indicators predating patch deployment. Follow CISA forensic triage steps if compromise detected. Federal agencies must patch by July 4, 2026 per CISA directive.
Source: CISA • Published: 2026-07-01
Relative path traversal vulnerability in keyhint option in repomd.xml parsing of libzypp before 17.38.12 (CVSS 8.4). Attackers can exploit this to write files outside intended directories during package repository operations.
Update libzypp to version 17.38.12 or later immediately. Review repository configurations for suspicious keyhint values. Audit systems for unexpected file modifications in system directories. Verify integrity of installed packages on affected systems.
Source: NVD • Published: 2026-07-02
Remote code execution in vLLM framework via malicious video link sent to API. Enables full server takeover including arbitrary command execution, data exfiltration, and lateral movement. Affects organizations exposing video models for user input.
Update vLLM to version 0.14.1 or later immediately. Implement strict input validation for video URLs including allowlist of trusted domains. Deploy sandboxing for video processing. Monitor for suspicious command execution and network connections from vLLM processes. Consider disabling video model endpoints until patched.
Source: GitHub • Published: 2026-04-20
Supply chain compromise of LiteLLM Python package on PyPI affecting tens of thousands of AI environments. TeamPCP threat actor published backdoored versions 1.82.7 and 1.82.8 using credentials stolen from prior Trivy compromise. Malicious packages available for 2+ hours with 3 million daily downloads creating significant victim base.
Immediately audit all dependencies for compromised LiteLLM versions 1.82.7 and 1.82.8. Replace with clean version from official source. Rotate all credentials and API keys that may have been exposed. Scan systems for backdoors and persistence mechanisms. Implement dependency pinning and hash verification. Review PyPI publish workflows for credential exposure.
Source: GTIG • Published: 2026-03-24
Shai-Hulud/Megalodon two-wave AI supply chain attack by TeamPCP. Wave 1: Self-propagating worm compromised 172 packages across 404 versions targeting Mistral AI, TanStack, Guardrails AI, UiPath. Wave 2 exploited developer toolchain within 3 weeks. Demonstrates sophisticated multi-layer supply chain targeting.
Audit all npm and PyPI dependencies from April-May 2026 against known malicious package list. Implement dependency integrity verification and artifact signing. Deploy SBOM analysis to detect compromised packages. Review developer toolchain security including CI/CD pipelines. Implement MAESTRO-aligned threat models including developer toolchain integrity and agentic AI attack surface.
Source: GTIG • Published: 2026-05-15
MCP security identified as urgent problem in AI enterprise systems. MCP servers provide AI agents access to databases, source code, email, cloud APIs, and production systems creating massive attack surface. MCP 2026-07-28 specification addresses security vulnerabilities but introduces new risks. Joint US government guidance issued on MCP security and prompt injection in tool-integrated LLM agents.
Treat every MCP server as hostile until proven otherwise. Implement gateway architecture with OAuth 2.1 authentication. Apply strict scoping to MCP server permissions. Deploy sandboxing for MCP server execution. Implement comprehensive logging and monitoring of MCP interactions. Review and validate parameter inputs and execution environment context. Follow joint government guidance on MCP security.
Source: OWASP • Published: 2026-06-15
Ransomware evolution shows 77% of 2025 intrusions involved data exfiltration with encryption, with trend accelerating in 2026 toward exfiltration-only attacks. Groups pivoting to DDoS-as-a-Service, insider recruitment using native English speakers, and gig worker exploitation due to decreased ransom payments. The Gentlemen group emerged Q3 2025 and became second most active with 182 Q1 2026 victims.
Implement data loss prevention to detect exfiltration before encryption. Deploy network segmentation to limit ransomware spread. Enhance insider threat detection programs focusing on recruitment indicators. Implement zero-trust architecture. Maintain offline backups with regular testing. Monitor for DDoS precursor activity. Prioritize vulnerability patching as 77% of initial access via known CVEs, stolen credentials, and phishing.
Source: GTIG • Published: 2026-07-01
Cybernews discovered 8.3TB publicly exposed database containing 24 billion credential records from 36 sources including Telegram channels, breach compilations, infostealer logs, and live server exports. More heavily weighted toward fresh infostealer logs than previous mega-dumps creating higher risk of valid credentials.
Assume credentials compromised and implement forced password reset campaigns. Deploy credential monitoring services to detect exposed organizational credentials. Accelerate multi-factor authentication rollout across all systems. Implement passwordless authentication where possible. Deploy compromised credential checking at login. Monitor for credential stuffing attacks. Review privileged account access. Educate users on password managers and unique passwords.
Source: Web • Published: 2026-06-25
Salt Typhoon Chinese APT compromised 200+ organizations across 80 countries including at least 8 US telecom providers (FBI, Verizon, AT&T, Lumen). Over 1 million people's metadata stolen. FBI confirmed in February 2026 threats remain active and ongoing. Norway and Singapore disclosed breaches. Senator Cantwell stated group may still be inside US networks. Experts believe attack not fully remediated.
Patch edge devices aggressively prioritizing VPN appliances, firewalls, and routers - Salt Typhoon initial access consistently exploits known CVEs per FBI February 2026 guidance. Conduct comprehensive network hunt for persistence mechanisms. Deploy enhanced logging and monitoring for telecom infrastructure. Implement network segmentation to limit lateral movement. Review and harden edge device configurations addressing basic security errors. Telecom providers must take enhanced protective action per Congressional oversight recommendations.
Source: CISA • Published: 2026-07-01
Volt Typhoon Chinese APT pre-positioned in U.S. critical infrastructure for potential disruption operations. ODNI 2026 Annual Threat Assessment highlights persistent cyber threats exhibiting tactics beyond traditional espionage. Demonstrates growing scope and sophistication of China's cyber capabilities targeting critical infrastructure.
Conduct immediate threat hunt for Volt Typhoon indicators of compromise in critical infrastructure networks. Implement enhanced monitoring and logging for OT/ICS environments. Deploy network segmentation between IT and OT networks. Review and restrict access to critical infrastructure control systems. Implement enhanced authentication for infrastructure access. Coordinate with FBI and CISA for threat intelligence sharing. Develop and test incident response plans for infrastructure disruption scenarios.
Source: CISA • Published: 2026-07-01
Department of Homeland Security HSIN platform breached late May to early June 2026 by unauthorized third party. HSIN used for strategic coordination and information sharing among security stakeholders. Potential exposure of security planning and coordination data for World Cup events. Under investigation.
DHS conducting forensic investigation to determine breach scope. All HSIN users should rotate credentials immediately. Review shared security plans for potential compromise. Implement enhanced authentication for HSIN access. Assess operational security impact for World Cup and other planned events. Deploy additional monitoring for HSIN platform. Consider temporary alternative secure communication channels for sensitive coordination.
Source: CISA • Published: 2026-07-01
CVE-2026-22778: Remote code execution in vLLM framework via malicious video link sent to API. Enables full server takeover including arbitrary command execution, data exfiltration, and lateral movement. Affects organizations exposing video models for user input.