Robust and Energy Efficient Malware Security for Robot Cyber-Physical Systems
Building scaled-down, distributed machine learning models for static malware detection on robots
Enabling each robot to secure itself against malicious software.The unique framework allows edge-level intelligence using scaled machine-learning models to ensure security. The first step in this endeavour was to create the first-of-its-kind dataset for malware and good software for robots, The RoboMal Dataset.
The RoboMal dataset, available at this repository!
Adversarial defense builds robustness into these models, while zero-shot learning allows such models to mature with time to learn previously unseen classes of malware. The overview of this thrust is available at ICCPS2022.
The adversarial defense pipeline for robustness in robot security.