PREVIEW. This is how it will look in an inbox; it was not sent to you. Each real email carries its own unsubscribe link and Keryn's postal address. KERYN HEADLINES Daily, Tuesday 27 October 2026 Retrospective Covers 26 October 2026 00:00 to 27 October 2026 00:00 UTC Web: https://keryn.dev/headlines/daily/2026-10-27/ TOP DEVELOPMENTS [1] litellm 1.107.0 Compromised on PyPI and Docker: Credential Stealer and Hourly Remote Code Loader StepSecurity · Incident · https://stepsecurity.io/blog/litellm-1-107-0-compromised-pypi-backdoored-wheels-credential-stealer [2] Cross-request KV cache exposure in the GLM-5.3 Workers AI model, and our remediation with SGLang Cloudflare: Security · Research · https://blog.cloudflare.com/kv-cache-exposure-remediation [3] PHP2Shell security flaws now sucessfully exploited in the wild BleepingComputer · Incident · https://bleepingcomputer.com/news/security/php2shell-security-flaws-now-sucessfully-exploited-in-the-wild SAST [4] ElectroNG is now open source! Doyensec · Analysis · https://blog.doyensec.com/2026/10/27/electrong-oss.html SCA [5] Software Supply Chain Security Solution Brief Endor Labs · Analysis · https://endorlabs.com/learn/software-supply-chain-security DAST [6] Former pentester used ARTEX AI to steal 8.2 million data records BleepingComputer · Analysis · https://bleepingcomputer.com/news/security/former-pentester-used-artex-ai-to-steal-82-million-data-records AGENTIC SDLC [7] Your Wish Is My Commit: Over-Compliant Refactoring in AI Coding Agents arXiv · Research · https://arxiv.org/abs/2610.36346 [8] Context Is a Pipeline, Not a Window: Information Attrition in LLM Agent Harnesses arXiv · Research · https://arxiv.org/abs/2610.36114 [9] Reva: Real-Time Vulnerability Detection through Contextual Probing of Large Language Models arXiv · Research · https://arxiv.org/abs/2610.35829 REMEDIATION [10] VU#128947: COVESA vsomeip excessive memory allocation in reliable (TCP) endpoint buffer handling CERT/CC · Analysis · https://kb.cert.org/vuls/id/128947 [11] Critical 'HTTP2Shell' Brotli vulnerability triggers emergency patches BleepingComputer · Analysis · https://bleepingcomputer.com/news/security/critical-http2shell-brotli-vulnerability-triggers-emergency-patches [12] How Absorption Rate Reveals Vulnerability Management Health Cloud Security Alliance · Analysis · https://cloudsecurityalliance.org/articles/how-absorption-rate-reveals-vulnerability-management-health TRUST [13] Atomic Security Concerns: Towards a Foundational Language for Security Knowledge Representation arXiv · Research · https://arxiv.org/abs/2610.35782 [14] Governing Enterprise AI Traffic at Scale: Design and Operation of a Policy-Driven, Multi-Vendor LLM Gateway on Envoy arXiv · Research · https://arxiv.org/abs/2610.35509 [15] From Representations to Computational Trajectories: Intrinsic Fingerprinting for Large Language Model Provenance Verification arXiv · Research · https://arxiv.org/abs/2610.34675 MODELS [16] SpecGuard: Detecting Multi-Turn LLM Jailbreaks via Specification-Based Runtime Monitoring arXiv · Research · https://arxiv.org/abs/2610.36258 [17] V$^3$AST-Bench: Verifiable, Valid, and Versatile Safety Evaluation for Coding Agents arXiv · Research · https://arxiv.org/abs/2610.35188 [18] The Target Agent Is Its Own Best Critic but a Poor Proposer: Self-Improved Red Teaming arXiv · Research · https://arxiv.org/abs/2610.34965 AI SECURITY [19] Log-To-Leak: Prompt Injection Attacks on Tool-Using LLM Agents via the Model Context Protocol arXiv · Research · https://arxiv.org/abs/2610.36108 [20] Then there is prompt injection, a malicious email, ...: The Wild Wide West of OpenClaw Agentic System Security arXiv · Research · https://arxiv.org/abs/2610.35451 [21] Scope-Gap Injection: Exploiting Evaluator Channel Blind Spots in the OpenClaw Agent Security Competition arXiv · Research · https://arxiv.org/abs/2610.34777 ADVISORIES IN THIS WINDOW 58 npm advisories · 58 critical 7 PyPI advisories · 7 critical All advisories: https://keryn.dev/headlines/advisories/ Titles, links and short excerpts only; every item links to its source. Preview: the unsubscribe link and Keryn's postal address are added to each email when it is sent.