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 14 of 2026 · Retrospective · Covers 24 March 2026 11:00 to 31 March 2026 11:00 UTC

Read on the web

Top developments

Compromised axios npm package delivers cross-platform RAT

Datadog · Incident · SCA

CISA: New Langflow flaw actively exploited to hijack AI workflows

BleepingComputer · Incident · AI Security

CISA orders feds to patch actively exploited Citrix flaw by Thursday

BleepingComputer · Advisory · Remediation

Backdoored Telnyx PyPI package pushes malware hidden in WAV audio

BleepingComputer · Incident · SCA

Tracking TeamPCP: Investigating Post-Compromise Attacks Seen in the Wild

Wiz · Incident · SCA

SAST

Software Vulnerability Detection Using a Lightweight Graph Neural Network

arXiv · Research · SAST

Attesting LLM Pipelines: Enforcing Verifiable Training and Release Claims

arXiv · Research · Models, SAST, SCA

Reachability Across the NL/PL Boundary: A Taxonomy-Driven Dataflow Model for LLM-Integrated Applications

arXiv · Research · SAST

VulnScout-C: A Lightweight Transformer for C Code Vulnerability Detection

arXiv · Research · SAST

SCA

GHSA-69fq-xp46-6x23 / CVE-2026-33634: Trivy ecosystem supply chain was briefly compromised

GitHub Advisory Database · Advisory · SCA

Popular LiteLLM PyPI package backdoored to steal credentials, auth tokens

BleepingComputer · Incident · SCA

New Torg Grabber infostealer malware targets 728 crypto wallets

BleepingComputer · Analysis · SCA

Wherefore Art Thou? Provenance-Guided Automatic Online Debugging with Lumos

arXiv · Research · SCA

DAST

Detecting and Mitigating Flakiness in REST API Fuzzing

arXiv · Research · DAST

Red-MIRROR: Agentic LLM-based Autonomous Penetration Testing with Reflective Verification and Knowledge-augmented Interaction

arXiv · Research · Agentic SDLC, DAST

Multi-target Coverage-based Greybox Fuzzing

arXiv · Research · DAST

Towards Leveraging LLMs to Generate Abstract Penetration Test Cases from Software Architecture

arXiv · Research · DAST

Agentic SDLC

Auto mode for Claude Code

Simon Willison · Analysis · Agentic SDLC

OpenAI Patches ChatGPT Data Exfiltration Flaw and Codex GitHub Token Vulnerability

The Hacker News · Advisory · Agentic SDLC, Remediation, SCA

Debt Behind the AI Boom: A Large-Scale Empirical Study of AI-Generated Code in the Wild

arXiv · Research · Agentic SDLC

Python Vulnerability Lookup

Simon Willison · Analysis · Agentic SDLC

Remediation

TP-Link warns users to patch critical router auth bypass flaw

BleepingComputer · Advisory · Remediation

Hackers exploiting critical F5 BIG-IP flaw in attacks, patch now

BleepingComputer · Analysis · Remediation

Policy-Driven Vulnerability Risk Quantification framework for Large-Scale Cloud Infrastructure Data Security

arXiv · Research · Remediation

Open VSX Bug Let Malicious VS Code Extensions Bypass Pre-Publish Security Checks

The Hacker News · Research · Remediation, SCA

Trust

Internet Yiff Machine: We hacked 93GB of "anonymous" crime tips

Ars Technica: Security · Incident · Trust

Beyond Corner Patches: Semantics-Aware Backdoor Attack in Federated Learning

arXiv · Research · Models, Trust

RAD-AI: Rethinking Architecture Documentation for AI-Augmented Ecosystems

arXiv · Research · Trust

"What Did It Actually Do?": Understanding Risk Awareness and Traceability for Computer-Use Agents

arXiv · Research · Trust

Models

MITRE ATLAS (releases) v5.5.0

MITRE · Release · AI Security, Models

Trojan-Speak: Bypassing Constitutional Classifiers with No Jailbreak Tax via Adversarial Finetuning

arXiv · Research · Models

Mitigating Backdoor Attacks in Federated Learning Using PPA and MiniMax Game Theory

arXiv · Research · Models

ORACAL: A Robust and Explainable Multimodal Framework for Smart Contract Vulnerability Detection with Causal Graph Enrichment

arXiv · Research · Models

AI Security

Covert Visual Prompt Injection against Commercial Multimodal Large Language Models

arXiv · Research · AI Security

Evaluating Privilege Usage of Agents with Real-World Tools

arXiv · Research · AI Security

Kill-Chain Canaries: Stage-Level Tracking of Prompt Injection Across Attack Surfaces and Five Production LLMs

arXiv · Research · AI Security

Evaluating Prompt Injection Defenses for Educational LLM Tutors: Security-Usability-Latency Trade-offs

arXiv · Research · AI Security, Models

Advisories in this window

  • 221 npm advisories · 98 critical · 53 high · 58 moderate · 12 low
  • 93 Go advisories · 9 critical · 28 high · 53 moderate · 3 low
  • 65 Packagist advisories · 3 critical · 26 high · 32 moderate · 4 low
  • 36 PyPI advisories · 6 critical · 16 high · 9 moderate · 5 low
  • 24 Maven advisories · 5 critical · 13 high · 5 moderate · 1 low
  • 13 NuGet advisories · 1 critical · 6 high · 4 moderate · 2 low
  • 11 crates.io advisories · 2 critical · 7 high · 2 moderate
  • 5 RubyGems advisories · 2 high · 1 moderate · 2 low
  • 2 Multi-ecosystem advisories · 1 moderate · 1 low
  • 1 GitHub Actions advisory · 1 critical
  • 1 Hex advisory · 1 high

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.