Issues

[new-cve] CVE-2026-108957 — A security flaw has been discovered in TinyLLaVA TinyLLaVA_Factory up to 1abc6ac

GenAI-Security-Project/crosswalk#296 · opened

on Oct 19, 2026

[new-research] arXiv:2610.22578 — Model-as-a-Trojan: Self-Activating Covert DNS Exfiltration Baked Into the Weight

GenAI-Security-Project/crosswalk#295 · opened

on Oct 19, 2026

[new-research] arXiv:2610.22697 — What It Is, Not What It Says: Securing Agentic Browsing Interactions with Web Se

GenAI-Security-Project/crosswalk#294 · opened

on Oct 19, 2026

[new-research] arXiv:2610.22722 — The LLM-Attacker: A Unified Red-Team Framework for Privacy Evaluation of Split a

GenAI-Security-Project/crosswalk#293 · opened

on Oct 19, 2026

[new-research] arXiv:2610.22828 — GRAD: Graph-Structured Antibody Defense for LLM Agents

GenAI-Security-Project/crosswalk#291 · opened

on Oct 19, 2026

[new-research] arXiv:2610.22800 — Veil-Then-Reveal: Protecting PII Privacy via Reversible Pseudo-Anonymization in

GenAI-Security-Project/crosswalk#292 · opened

on Oct 19, 2026

[new-research] arXiv:2610.22899 — Provable Authority: A Protocol for Verifiable Authority Resolution and Control A

GenAI-Security-Project/crosswalk#289 · opened

on Oct 19, 2026

[new-research] arXiv:2610.22874 — Identify, Classify, Act: Per-Instruction Controlled Compliance for LLMs over Ext

GenAI-Security-Project/crosswalk#290 · opened

on Oct 19, 2026

[new-research] arXiv:2610.22959 — SAFE-Mem: Secure Agentic Memory with Fine-Grained Authority and Governance-Aware

GenAI-Security-Project/crosswalk#288 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23030 — ECHO: Replacing the Compliance Proxy with Response Harmfulness in Optimization-B

GenAI-Security-Project/crosswalk#287 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23077 — When Approval Becomes Execution: Authorization-Grounded Workflow Attacks on Tool

GenAI-Security-Project/crosswalk#285 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23035 — VulnSAGE: Adjudicating SAST Reports by Connecting LLM Reasoning to Program Struc

GenAI-Security-Project/crosswalk#286 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23078 — Cryptographically Secure Agent-to-Agent Communication in Multi-Agent Systems

GenAI-Security-Project/crosswalk#284 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23109 — Trojan Horse: Automated Repository Poisoning against Coding Agents

GenAI-Security-Project/crosswalk#283 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23233 — AgentLoupe: Kernel-Resident Semantic Causal Labels for AI Agent Observability

GenAI-Security-Project/crosswalk#282 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23254 — Ignored, Not Refused: What Low Prompt-Injection Success Rates Hide in Small Tool

GenAI-Security-Project/crosswalk#281 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23303 — MobileJailBench: Red Queen Jailbreak Benchmarking for Mobile Agents

GenAI-Security-Project/crosswalk#280 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23307 — MobileInject: Evaluating Reasoning-to-Action Security of MLLM-Based Mobile Agent

GenAI-Security-Project/crosswalk#279 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23362 — Monitor-Validity Failures: LLM Safety Evidence Paths at Evolving Framework Bound

GenAI-Security-Project/crosswalk#278 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23601 — Token-Level Attribution of Multi-Mode LLM Watermarks: Optimal Rule and Belief We

GenAI-Security-Project/crosswalk#277 · opened

on Oct 19, 2026

[new-research] arXiv:2610.23656 — ABF-Bench: A Verifiable Benchmark for Agentic Blockchain Forensics

GenAI-Security-Project/crosswalk#276 · opened

on Oct 19, 2026

[new-cve] CVE-2026-108593 — 9router 0.4.1 through 0.5.99 contains a configuration injection vulnerability in

GenAI-Security-Project/crosswalk#275 · opened

on Oct 12, 2026

[new-cve] CVE-2026-107295 — Pydantic AI is a Python agent framework for building applications and workflows

GenAI-Security-Project/crosswalk#274 · opened

on Oct 12, 2026

[new-cve] CVE-2026-107294 — Pydantic AI is a Python agent framework for building applications and workflows

GenAI-Security-Project/crosswalk#273 · opened

on Oct 12, 2026

[new-cve] CVE-2026-107292 — Pydantic AI is a Python agent framework for building applications and workflows

GenAI-Security-Project/crosswalk#271 · opened

on Oct 12, 2026

Navigation Menu

Repository navigation

Issues

[new-cve] CVE-2026-108957 — A security flaw has been discovered in TinyLLaVA TinyLLaVA_Factory up to 1abc6ac

[new-research] arXiv:2610.22578 — Model-as-a-Trojan: Self-Activating Covert DNS Exfiltration Baked Into the Weight

[new-research] arXiv:2610.22697 — What It Is, Not What It Says: Securing Agentic Browsing Interactions with Web Se

[new-research] arXiv:2610.22722 — The LLM-Attacker: A Unified Red-Team Framework for Privacy Evaluation of Split a

[new-research] arXiv:2610.22828 — GRAD: Graph-Structured Antibody Defense for LLM Agents

[new-research] arXiv:2610.22800 — Veil-Then-Reveal: Protecting PII Privacy via Reversible Pseudo-Anonymization in

[new-research] arXiv:2610.22899 — Provable Authority: A Protocol for Verifiable Authority Resolution and Control A

[new-research] arXiv:2610.22874 — Identify, Classify, Act: Per-Instruction Controlled Compliance for LLMs over Ext

[new-research] arXiv:2610.22959 — SAFE-Mem: Secure Agentic Memory with Fine-Grained Authority and Governance-Aware

[new-research] arXiv:2610.23030 — ECHO: Replacing the Compliance Proxy with Response Harmfulness in Optimization-B

[new-research] arXiv:2610.23077 — When Approval Becomes Execution: Authorization-Grounded Workflow Attacks on Tool

[new-research] arXiv:2610.23035 — VulnSAGE: Adjudicating SAST Reports by Connecting LLM Reasoning to Program Struc

[new-research] arXiv:2610.23078 — Cryptographically Secure Agent-to-Agent Communication in Multi-Agent Systems

[new-research] arXiv:2610.23109 — Trojan Horse: Automated Repository Poisoning against Coding Agents

[new-research] arXiv:2610.23233 — AgentLoupe: Kernel-Resident Semantic Causal Labels for AI Agent Observability

[new-research] arXiv:2610.23254 — Ignored, Not Refused: What Low Prompt-Injection Success Rates Hide in Small Tool

[new-research] arXiv:2610.23303 — MobileJailBench: Red Queen Jailbreak Benchmarking for Mobile Agents

[new-research] arXiv:2610.23307 — MobileInject: Evaluating Reasoning-to-Action Security of MLLM-Based Mobile Agent

[new-research] arXiv:2610.23362 — Monitor-Validity Failures: LLM Safety Evidence Paths at Evolving Framework Bound

[new-research] arXiv:2610.23601 — Token-Level Attribution of Multi-Mode LLM Watermarks: Optimal Rule and Belief We

[new-research] arXiv:2610.23656 — ABF-Bench: A Verifiable Benchmark for Agentic Blockchain Forensics

[new-cve] CVE-2026-108593 — 9router 0.4.1 through 0.5.99 contains a configuration injection vulnerability in

[new-cve] CVE-2026-107295 — Pydantic AI is a Python agent framework for building applications and workflows

[new-cve] CVE-2026-107294 — Pydantic AI is a Python agent framework for building applications and workflows

[new-cve] CVE-2026-107292 — Pydantic AI is a Python agent framework for building applications and workflows