Software and Information Systems
Assistant Professor
Woodward Hall 333C
- Anomaly Detection aims to extract the invariant patterns of IoT systems’ normal behaviors from semantic information and training datasets. IoT systems’ abnormal behaviors are detected as violations or deviations of the learned patterns.
- IoT Vulnerability Discovery utilizes software testing tools such as formal verification and fuzzing to localize desinging and implementation flaws that cause security policy violations. Knowledge that is specific to the IoT domain helps to automate and optimize the procedure of vulnerability discovery.
- Adversarial Example Detection focuses on identifying malicious input samples of deep neural networks that are specially crafted with human-imperceptible perturbations. By exploiting adversarial examples’ limitations on transferability and robustness, their attack attempts can be either detected or defeated.
