Stefan Schmid

Repeated interactions with stable marriage algorithms can expose the private preferences of non-malicious participants, revealing significant vulnerabilities in widely used matching systems.

SecureClaw achieves a remarkable 0% attack success rate while preserving task utility, setting a new standard for LLM agent security.

Achieving up to 50% faster secure inference for LLMs without sacrificing accuracy could revolutionize the deployment of privacy-preserving AI systems.