- Host header validation flaw in Starlette affecting access control
- LiteLLM vulnerability chained with CVE-2026-42271 for RCE
ThreatNoir Afternoon Brief — November 16
Afternoon Review in IT Security — November 16, 2026
The threat landscape continues to evolve rapidly as artificial intelligence reshapes both attack and defense capabilities, while critical vulnerabilities in widely-deployed infrastructure remain actively exploited in the wild. Today's security briefing covers developments ranging from autonomous AI-driven exploitation to supply-chain compromises affecting Russian software developers.
Google: AI Agent Turned a Security Patch Into a Working Exploit With No Human Help
Researchers at Google have demonstrated a significant shift in the threat model surrounding vulnerability exploitation. An AI agent successfully transformed a security patch into a fully functional exploit without any human intervention in the process. Source: Google: AI agent turned a security patch into a working exploit with no human help. This development underscores the accelerating capability of machine learning systems to reverse-engineer security mitigations and weaponize them at scale, raising critical concerns about the future timeline for vulnerability disclosure and patch deployment strategies.
Scans for LiteLLM and Starlette Vulnerabilities Detected in Honeypots
Security researchers have begun observing active scanning attempts targeting vulnerabilities in LiteLLM and Starlette frameworks, specifically CVE-2026-42271 and CVE-2026-48710. The reconnaissance activity, detected over the weekend as requests to the /mcp-rest/test/tools/list endpoint, indicates that threat actors have already begun attempting to exploit these flaws in production environments. Source: Scans for LiteLLM/Starlette (CVE-2026-42271, CVE-2026-48710), (Mon, Nov 16th). Security researchers have documented how these two vulnerabilities can be chained together to achieve unauthenticated remote code execution, making them particularly attractive targets for attackers seeking initial access to cloud-based infrastructure.
Recent UnrealIRCd Vulnerability Exploited in Attacks
A high-severity vulnerability in UnrealIRCd, tracked as CVE-2026-114713, is being actively exploited in the wild just days after its disclosure. The flaw permits attackers to execute arbitrary commands with server and IRCOp-level privileges, providing a direct pathway to infrastructure compromise. Source: Recent UnrealIRCd Vulnerability Exploited in Attacks. Organizations operating IRC infrastructure should prioritize immediate patching to prevent unauthorized command execution and potential lateral movement within their networks.
BO Team Exploits Log4Shell in VMware vCenter, Compromises Russian Firm's Code-Signing Server
A months-long investigation by Kaspersky's ICS CERT team has revealed a sophisticated supply-chain attack in which the threat actor BO Team exploited the Log4Shell vulnerability (CVE-2021-44228) in VMware vCenter to gain initial access to a Russian software developer's infrastructure. The attackers subsequently compromised the organization's code-signing server, creating a significant risk for downstream customers who may have received trojanized software updates. Source: BO Team Exploits Log4Shell in VMware vCenter, Compromises Russian Firm's Code-Signing Server. The investigation documented the deployment of multiple post-exploitation tools including BrockenDoor, Cobalt Strike, Mythic, and ZeronetKit, indicating a well-resourced adversary with sustained access to the victim's development infrastructure for an extended period.
Today's threat intelligence reinforces the critical importance of timely vulnerability patching, continuous monitoring for exploitation attempts, and heightened vigilance regarding supply-chain security in software development environments. Organizations should prioritize remediation of known critical vulnerabilities while simultaneously implementing detection mechanisms for active exploitation campaigns.
Sources & IOCs
Source articles and their indicators (defanged where appropriate). IP addresses, domains, URLs and file hashes are copied from the publisher's page by pattern matching. Malware names, CVE numbers and technique ids are picked by AI. No person reviews them.
- UnrealIRCd vulnerability
- Check Point products vulnerability
- Log4Shell vulnerability exploited in VMware vCenter
- Cobalt StrikeTool used by BO Team
- BrockenDoorMalware used by BO Team
- ZeronetKitMalware used by BO Team
- MythicTool used by BO Team
193.3.23.236Written defanged in the article text80.76.42.139Written defanged in the article text
lo1yaizaeghaevarooD1ahtaichootae.nofrillsdns.comCommand and control domain for backdoor scriptlo1yaizaeghaevarood1ahtaichootaz.nofrillsdns.comWritten defanged in the article text