Live intelligence

Latest AI updates

2 updates
Research

Securing Multi-Agent Systems by Design

Researchers propose response manipulation attacks (ReMA), which iteratively optimize malicious agent responses against multi-agent systems. They also introduce SecMAS, a secure aggregation rule that uses semantic similarity to select responses consistent with a sufficiently large subset of peers. Theoretical guarantees cover untargeted and targeted attacks, and evaluations show existing aggregation rules are vulnerable while SecMAS defends effectively with comparable utility.

multi-agent systemssecurityadversarial machine learning
Research

Logit Laundering: Evading LLM Data-Use Auditing via Inference-Time Logit Reshaping

The paper examines watermark-based data-use auditing for LLMs from an adversarial angle, positing a non-compliant provider that trains on unauthorized data and tries to hide it from auditors. It proposes Logit Laundering, a framework that keeps model parameters unchanged and instead manipulates inference-time logits through a logit reshaping module with a plug-and-play statistical bias estimator to suppress watermark-induced signals. A confidence-aware scaling module adaptively calibrates suppression strength based on next-token prediction confidence to trade off evasion against model utility. The authors report evaluations showing evasion of several state-of-the-art auditing methods while preserving downstream task performance, and argue for more robust auditing approaches.

LLM watermarkingtraining data auditingadversarial machine learning