Research
IoT and Network Security
Security of wireless communications among Internet of Things (IoT) Devices is threatened by adversaries that aim to disrupt the communication among these devices. In this research, we are interested in designing efficient communication mechanisms among IoT devices to maximize network throughput. In addition, in this research we work on fining efficient techniques for detection of malicious users to take countermeasure actions against them.
Cyber Physical System Security
Cyber Physical Systems(CPS) are systems where their operations can be modeled as three features of physical, cyber, and control. There is coupling among these features and the optimal coordination of these features results in the optimal performance of a CPS. There are many industrial, and health applications that can be modeled with CPS such as smart grids, vehicle platooning, etc. Similar to other wireless devices, the security of CPS is threatened by attackers. We are interested in studying the coupling between three features in CPS to guarantee stability and performance in CPS.
Machine Learning Security
Machine Learning techniques has been proven to be promising for learning various tasks and has been widely applied in various applications including but not limited to health, disease diagnosis, image classification, autonomous driving, security, text classification, and many other. However, it has recently been studies that these learning mechanisms are vulnerable against either adversarial trial data or test data also called adversarial examples. We are interested to designing secure machine learning algorithms that can be effectively applied in cybersecurity applications.
Online Machine Learning
Online machine learning algorithms have many applications including but not limited to online advertising, new customization, clinical trials, cybersecurity, optimal policing, etc. Online learning is a category of reinforcement learning where in the agent learns the best action to take as each state to maximize its gain in a time horizon. The agent at each step needs to trade off between exploration and exploitation. Online machine learning techniques have many variants and we are interested in designing and applying these mechanisms in all aspects of cybersecurity.
