Advancing Network Security, Machine Learning & Game Theory
Cyber security is a complex and challenging problem. The design of defense mechanisms to secure networks and systems has received immense attention from the research community for more than two decades. However, the problem is far from completely solved. In the GTCS Lab, we are exploring the applicability of game-theoretic, machine learning, and other advanced mathematical approaches to address critical security challenges in networks, cloud environments, and intelligent systems.
Current Research Areas
Machine Learning Testing
Evaluating the correctness, robustness, and reliability of machine learning systems to identify faults and ensure consistent behavior under changing conditions.
- Metamorphic Testing
- Mutation Testing
Agentic AI Security
Investigating threats, developing defenses, and evaluating benchmarks to strengthen the security and reliability of autonomous AI agents.
- Threat Analysis and Defense Mechanisms
- Security Benchmarks and Agent Evaluation
Interpretable Machine Learning
Explaining the predictions and internal logic of black-box models to establish trustworthiness and select critical features.
- Consistency Analysis
- DDoS Explanation and Feature Selection
ML Ensemble Frameworks
Combining multiple supervised and unsupervised classifiers to achieve robust, high-accuracy detection of network intrusions.
- Optimal Feature Selection
- Supervised & Unsupervised Models
Crowdsourcing
Designing multi-objective recommendation algorithms to optimize software development tasks for workers, employers, and providers.
- Task Recommendation Algorithms
- Self-Serving Stakeholder Modeling
Game Theory Defense (GIDA)
Developing a semi-autonomous Game Theory Inspired Defense Architecture that employs a "carrot and stick" strategy against attackers.
- Stochastic Games with Imperfect Info
- Honeynets & Dynamic Redirects