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

Weekly, week 34 of 2025 · Retrospective · Covers 12 August 2025 11:00 to 19 August 2025 11:00 UTC

Read on the web

Top developments

Microsoft August 2025 Patch Tuesday fixes one zero-day, 107 flaws

BleepingComputer · Advisory · Remediation

Amazon Q Developer: Secrets Leaked via DNS and Prompt Injection

Embrace The Red · Incident · Agentic SDLC, AI Security, SCA

Docker Hub still hosts dozens of Linux images with the XZ backdoor

BleepingComputer · Analysis · SCA

The Summer of Johann: prompt injections as far as the eye can see

Simon Willison · Research · Agentic SDLC, AI Security

Case Study: How LFX Insights and OSPS Baseline Validated GUAC’s Security in Under an Hour

OpenSSF · Research · SCA

SAST

MCP-Guard: A Multi-Stage Defense-in-Depth Framework for Securing Model Context Protocol in Agentic AI

arXiv · Research · AI Security, Models, SAST

Data and Context Matter: Towards Generalizing AI-based Software Vulnerability Detection

arXiv · Research · SAST

Enhancing GraphQL Security by Detecting Malicious Queries Using Large Language Models, Sentence Transformers, and Convolutional Neural Networks

arXiv · Research · SAST

Code Vulnerability Detection Across Different Programming Languages with AI Models

arXiv · Research · SAST

SCA

Beneath the Mask: Can Contribution Data Unveil Malicious Personas in Open-Source Projects?

arXiv · Research · SCA

Malicious PyPI and npm Packages Discovered Exploiting Dependencies in Supply Chain Attacks

The Hacker News · Research · SCA

Preventing Domain Resurrection Attacks

Python Package Index · Analysis · SCA

DAST

Trivial C# Random Exploitation

Doyensec · Analysis · DAST

Reducing False Positives with Active Behavioral Analysis for Cloud Security

arXiv · Research · DAST, Models

XAMT: Cross-Framework API Matching for Testing Deep Learning Libraries

arXiv · Research · DAST

Agentic SDLC

Data Exfiltration via Image Rendering Fixed in Amp Code

Embrace The Red · Analysis · Agentic SDLC, AI Security, Remediation

Amp Code: Invisible Prompt Injection Fixed by Sourcegraph

Embrace The Red · Analysis · Agentic SDLC, AI Security

Google Jules: Vulnerable to Multiple Data Exfiltration Issues

Embrace The Red · Analysis · Agentic SDLC

Remediation

Fortinet warns of FortiSIEM pre-auth RCE flaw with exploit in the wild

BleepingComputer · Analysis · Remediation

Microsoft Windows Vulnerability Exploited to Deploy PipeMagic RansomExx Malware

The Hacker News · Incident · Remediation

Plex warns users to patch security vulnerability immediately

BleepingComputer · Analysis · Remediation

Microsoft fixes Windows Server bug causing cluster, VM issues

BleepingComputer · Analysis · Remediation

Trust

Android's pKVM hypervisor earns SESIP Level 5 security certification

BleepingComputer · Analysis · Trust

Documenting Deployment with Fabric: A Repository of Real-World AI Governance

arXiv · Research · Trust

AI-Augmented CI/CD Pipelines: From Code Commit to Production with Autonomous Decisions

arXiv · Research · Trust

Rethinking Autonomy: Preventing Failures in AI-Driven Software Engineering

arXiv · Research · Trust

Models

CCFC: Core & Core-Full-Core Dual-Track Defense for LLM Jailbreak Protection

arXiv · Research · AI Security, Models

MAJIC: Markovian Adaptive Jailbreaking via Iterative Composition of Diverse Innovative Strategies

arXiv · Research · Models

RAJ-PGA: Reasoning-Activated Jailbreak and Principle-Guided Alignment Framework for Large Reasoning Models

arXiv · Research · Models

Involuntary Jailbreak: On Self-Prompting Attacks

arXiv · Research · Models

AI Security

MCPTox: A Benchmark for Tool Poisoning Attack on Real-World MCP Servers

arXiv · Research · AI Security

MCPXKIT: The Unified Toolkit for Analyzing Model Context Protocol Security

arXiv · Research · AI Security

MCPSecBench: A Systematic Security Benchmark and Playground for Testing Model Context Protocols

arXiv · Research · AI Security

Too Easily Fooled? Prompt Injection Breaks LLMs on Frustratingly Simple Multiple-Choice Questions

arXiv · Research · AI Security

Advisories in this window

  • 14 Maven advisories · 2 high · 9 moderate · 3 low
  • 9 npm advisories · 4 critical · 1 high · 3 moderate · 1 low
  • 9 Packagist advisories · 5 high · 4 moderate
  • 8 PyPI advisories · 1 high · 7 moderate
  • 8 Go advisories · 1 critical · 3 high · 4 moderate
  • 3 crates.io advisories · 1 high · 2 moderate
  • 3 RubyGems advisories · 2 critical · 1 moderate
  • 1 Swift advisory · 1 moderate
  • 1 GitHub Actions advisory · 1 critical

All advisories

Titles, links and short excerpts only; every item links to its source. Selection is automatic; the editor's note is AI-drafted and checked as described. Preview: the unsubscribe link and Keryn's postal address are added to each email when it is sent.