Vasisht Duddu

I am a Distinguished Postdoctoral Fellow at Vector Institute for Artificial Intelligence advised by Tim Rudner and Xi He. I received my Ph.D. and Master’s in Computer Science from University of Waterloo advised by N. Asokan, and my undergraduate from IIIT-Delhi.

Research: I work on trustworthy machine learning systems by grounding my research in systems security principles:

Assessing Risks: Define realistic threat models to evaluate risks to security, privacy, safety, fairness, transparency
Designing Defenses: Design defenses by identifying application-specific requirements (e.g., effective, efficient, robust)
Holistic System Analysis: Study unintended interactions and trade-offs of ML models with other system components
Technical Governance: Design mechanisms for accountability and regulatory compliance

Selected Publications:

Assessing Risks: MobiQuitous'20, CCS'23, ICML'25, PETS'26, ArXiv'26
Designing Defenses: S&P'24, CODASPY'25 (best paper), EACL Findings'26, ACL'26
Holistic System Analysis: CIKM'22, S&P'24 (distinguished paper), TMLR'25, USENIX SEC'26
Technical Governance: ESORICS'24, CODASPY'25, ArXiv'26

Selected Research Awards (full list):

Photo Credits: Joe Petrik