Four isometric layers rise from bottom left to top right. I Hardware: a chip whose gold enclave holds a red root of trust. II Systems: three virtual machines on a hypervisor; the middle one is gold, a confidential VM, and a red crack splits the one on the right, a vulnerability. III Models: a small neural network whose red inference path ends in a gold output node marked with a check. IV Agents: a gold agent block marked with the AI spark, linked to a tool isolated in a wireframe cage. Gold marks what is protected on each layer; one red thread, the chain of trust, rises from the root of trust through every gold element to the agent. I  Hardware II  Systems III  Models IV  Agents
Trustworthy computing, from hardware to AI agents.

SEPT Lab (Security, Privacy, and Trust) is a research lab in the Thomas Lord Department of Computer Science at the University of Southern California. SEPT Lab is led by Prof. Mengyuan Li.

SEPT Lab designs trustworthy computing environments by tightly integrating hardware security mechanisms with software systems, so that private data and programs stay protected on platforms from personal devices to cloud AI systems. Our research has two thrusts: uncovering, understanding, and defending against system and hardware vulnerabilities across cloud systems, CPUs, GPUs, and AI systems; and co-designing software and hardware for secure and efficient AI systems. Please refer to our publications to see examples of our research.

We also teach graduate courses on confidential computing, side channels, and systems security at USC.

We are looking for motivated PhD, MS, and undergraduate students; see how to join SEPT Lab.

SEPT Journey

Four research directions, from Hardware to Agents