Atul Prakash

Atul Prakash is a Professor in the Department of Electrical Engineering and Computer Science at the University of Michigan, serving as the Richard H. Orenstein Division Chair of Computer Science and Engineering. His research focuses on security and privacy of emerging systems, particularly in cyber-physical systems and efficient and secure machine learning. He has led groundbreaking work on IoT security, adversarial machine learning, and financial web security, with notable papers receiving awards such as the Distinguished Practical Paper Award (IEEE S&P 2016) and Best Paper Award (SECDEV 2018).

• University of Michigan, College of Engineering
• Research Interests: Security of ML, IoT/CPS, Adversarial Examples, Financial Web Security

Recent publications highlight trends in adversarial machine learning, deepfake detection, and IoT vulnerabilities, with applications in large language models, computer vision, and mobile payment systems. His work spans both theoretical and practical aspects of cybersecurity, often involving collaborations with institutions like Microsoft Research and Stony Brook University.

• 2024: NeurIPS (Structured Sparsity in LLMs), ACL (LLM Guard-Rail Attacks), WACV (Diffusion Deepfake Detection)
• 2023: ICML (OoD Explanations), ICLR (Coreset Pruning)
• 2022: EuroS&P (Physical Adversarial Examples), USENIX (Mobile App Geodifferences), IEEE (Deepfake Analysis)

Scientific awards include:

• IEEE S&P 2016: Distinguished Practical Paper Award
• SECDEV 2018: Best Paper Award

Students and advisees include current Ph.D. candidates like Ryan Feng and Neal Mangaokar, and alum such as Renuka Kumar (Cisco), Kevin Eykholt (IBM), and Patrick McDaniel (Penn State). He actively mentors undergraduate and graduate students, with notable alumni at Google, Netflix, and Microsoft.

Labs and teams: Leads the Security of ML and IoT research group at the University of Michigan, collaborating with NSF-funded projects (Grants #1318722, #1646392, #17408977).