NeurIPS 2026 reviews are up: reviews, scores, rebuttals and meta-reviews for 7,877 papers. See the highest rated
acceptodds: the papers of ICLR 2027
100 results for “keyword:"agent security"” in ICLR 2027. 22 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]From Declared Capabilities to Behavior: Detecting Malicious Agent Skills Through Consistency Analysis▲9.2k2.67
[PDF]MELT:A Dependency-Aware Defense against Memory Poisoning Attacks Based on Taint Analysis▼4.5k2.67
[PDF]When Correct Plans Lead to Wrong Actions: Multi-Source Evidence Fabrication Against LLM Agents▼4.3k2.00
[PDF]ActionPatch: Distilling Adversarial Instructions into Visual Patches for GUI Agent Hijacking▲502.75
[PDF]Can Useful Work Be the Attack? CUCKOO SKILLS: PARASITIC TOKEN-COST ATTACKS ON LLM AGENTS▲3.0k2.33
[PDF]CoSA: Characterizing the Conflict of Skills Attack in Real-World LLM Agent Skill Ecosystems▼5.8k2.33
[PDF]Runtime Monitoring of Distributed Backdoors in Multi-Agent LLM Systems: Local Detection, Encoding Changes, and Blocking▼5.9k2.50
[PDF]Endogenous Byzantine Faults: Measuring and Containing Fault Contagion in LLM Multi-Agent Systems▲3.0k2.50
[PDF]ArenaWeaver: Automatic Generation of Environments and Attacks for Prompt Injection Evaluation at Scale▼4.5k2.67
[PDF]AgentHoneypot: Benchmarking Machine-Readable Content Injection and Provenance-Aware Authorization in Web Agents▼4.3k2.00
[PDF]Policy-State Fragmentation in Multi-Agent LLM Systems: Analyzing Binding Continuity and Runtime Enforcement▼1.2k1.67
[PDF]DecodingTrust-Agent Platform (DTap): A Controllable and Interactive Red-Teaming Platform for AI Agents▲44.00
[PDF]From Untrusted Content to Accepted Instructions: Security Risks of Context Compaction in LLM Agents▼3.7k2.25
[PDF]AutoDojo: A Generative Benchmark for Evaluating Prompt Injection Defenses in LLM Agents▼4.3k2.00
[PDF]Divide and Inject: Can Agents Reconstruct an Indirect Prompt Injection from Fragments?▲3.0k2.33
[PDF]Self-Reflection Fine-Tuning: Enhancing Agent Security against Prompt Injection Attacks from Failure Experience▲4012.67
[PDF]Can Agents Trust Their Skills? Uncovering Unsafe Chains of Trust in Skill-Based LLM Agents▼5.8k2.33
[PDF]From Evidence to Effect: Authority Semantics and Runtime Infrastructure for Stateful Agents▼1.2k1.67
[PDF]Hiding in Plain Sight: Decoupling Pretext from Actuation for Skill Poisoning in LLM Agents▼4.5k2.67
[PDF]TrojanAgent: A Malicious Agent Bypassing Guardrails to Execute Dangerous Actions in Multi-Agent Systems▼5.8k2.33
[PDF]WardenGuard and Warden-Bench: A Multichannel Defense and Benchmark for Indirect and Agentic Prompt Injection▼4.3k2.00
[PDF]The Announcement, Not the Attack: What LLM Safety Judges Answer and What Their Hidden States Hold▲5.4k3.00
[PDF]Beyond Centrality: Position-Decoupled Verification for Finite-Horizon Attacks in LLM Multi-Agent Systems▼4.3k2.00
Withdrawn or desk rejected (22)
[PDF]SPILLWAY: Separating Target Admission from Sanitization to Defend Tool-Using Agents against Prompt InjectionWithdrawn2.00
[PDF]Adaptive-Attack Robustness of Prompt-Injection Detectors for Agentic Tool-Response ContentDesk rejected
[PDF]Harmless Here, Harmful There: Safety and Privacy Inversion of Agent Memory Across Model MigrationWithdrawn1.33
[PDF]HelpInject: Reframing Adversarial Goals as Helpful Solutions for Indirect Prompt Injection in LLM AgentsWithdrawn1.67