Karthik Pattabiraman
Papers
12
Total Citations
229
H-Index
7
About
Karthik Pattabiraman is a prominent researcher whose work sits at the critical intersection of system security, fault tolerance, and autonomous systems. His research has made significant contributions to two primary domains: the resilience of machine learning systems and the security of robotic vehicles (RVs). Pattabiraman's early high-impact work includes TensorFI (2018, 63 citations), a configurable fault injection framework for TensorFlow applications that has become a foundational tool for evaluating ML system reliability in safety-critical contexts. This work laid the groundwork for his broader investigations into error-resilient machine learning, particularly for autonomous vehicles and industrial robotics. A substantial body of his research targets the security vulnerabilities of robotic vehicles against physical sensor attacks. His suite of interconnected projects — including "Out of Control" (45 citations), PID-Piper (41 citations), and DeLorean — progressively advanced the state of the art in attack detection, diagnosis, and recovery for RVs subjected to GPS spoofing and sensor tampering. More recently, SpecGuard and SwarmFuzz have extended this work to drone swarms and mission-specification-aware recovery. Pattabiraman's research addresses urgent real-world challenges in autonomous systems security, making it essential reading for researchers working on dependable robotics, adversarial resilience, and safety-critical AI deployments.
Research Focus
Key Achievements
Top Papers
- 1TensorFI: A Configurable Fault Injector for TensorFlow Applications63 citations · 2018
- 2Out of control45 citations · 2019
- 3PID-Piper: Recovering Robotic Vehicles from Physical Attacks41 citations · 2021
- 4
- 5Poster8 citations · 2022
- 6
- 7SwarmFuzz: Discovering GPS Spoofing Attacks in Drone Swarms7 citations · 2023
- 8
- 9
- 10