Directory profile for Dr. Sayanton Dibbo

Dr. Sayanton Dibbo’s research mission is to identify and systematically study adversarial and privacy threats to AI/ML systems under different realistic setups/assumptions, including Generative AI models, and develop innovative defense frameworks/tools to mitigate AI/ML vulnerabilities. His primary objective is to improve the AI/ML systems’ robustness to ensure the AI/ML systems are more secure and trustworthy. AI/ML cyberattacks (i.e., adversarial and privacy attacks) aim to make the model/system vulnerable by generating incorrect predictions or allowing the AI/ML systems to leak/infer sensitive private training data. This scenario is even more threatening in the case of generative AI (GenAI) models. As a result, innovative and cutting-edge defense tools are necessary to ensure secure and trustworthy AI/ML computations by improving the robustness of the AI/ML systems against cyberattacks.

Dr. Dibbo’s research integrates AI/ML and Security/Privacy domains. In particular, his research investigates the impact of different security and privacy threats on different data modalities, including images, texts, tabular, and audio data. His research vision is to develop novel tools that can mitigate adversarial and privacy attacks targeting AI/ML systems involving various data modalities. Dr. Dibbo’s outstanding research has been published in top-tier AI/ML and Cybersecurity conferences and journals, including USENIX Security, IEEE Computer Security Foundation (CSF), IEEE Secure and Trustworthy ML (SaTML), IEEE Transactions on Dependable and Secure Computing, ACM Conference on Computer and Communications Security (CCS), and European Conference on Computer Vision (ECCV). He has also served as a reviewer and technical program committee member in several top ACM/IEEE/Springer conferences and journals.