Autonomous and Distributed Security
Game and Control Theoretic Approach
We develop and analyze quantitative approaches to decision, control, and resource allocation problems arising in the context of autonomous and distributed security. Despite the recent surge of interest in the subject, most of the existing research efforts are heuristic in nature. Utilizing game and -optimal- control theories as a basis, our objective is to develop formal mathematical frameworks to model and solve autonomous security problems. We believe that such mathematical abstractions are useful for generalization of problems, combining the existing ad-hoc schemes under a single umbrella, and facilitating future research. We have successfully applied this approach to topics such as optimal deployment and configuration of malware filters, strategies for (malware) epidemic removal, and allocation of (system administrator) resources for detection and response.
Distributed Intrusion Detection and Response
With the objective of developing a scalable, robust, and effective intrusion detection and response architecture, we investigate an approach combining P2P communication schemes, machine learning techniques, and software agent technologies, which we call "autonomous security". Specifically, we envision distributed and agent-based methods facilitating collaborative decision making and response. This research effort is a collaboration with researchers in Dai-Lab, TUB where I take part in management of the security group and co-supervise students.
Secure and Intelligent Services
There is a lot of room for improvement in the area of security services such as authentication and authorization where the current situation has become a burden for users leading to not only bad usability but also weak security. We target to address, on the one hand, the password-based authentication problems through biometric authentication, especially using haptics-based systems. On the other hand, we investigate improvements to decentralized identity management, specifically OpenID.
| Top of page |
Distributed Services and Social Networks
Services provided over distributed systems, especially the Internet, is an increasingly popular research field that is related to decision and control, distributed machine learning, and networking. Analysis and development of scalable and distributed algorithms for optimal allocation of resources based on pricing, optimization, and game theory constitute one specific direction. Another active research area is the organization and intelligent routing of information utilizing machine learning, information retrieval, and social networking. In a recent project (internal codename Spree), we investigate the underlying principles and develop a prototype of a social search system for information exchange between users of a community. By its very nature, this project integrates diverse areas including networking, machine learning, human-computer interface design, and usability.
| Top of page |
Optimization and Control of Networks
We analyze, develop, and implement distributed and scalable pricing, control, and optimization schemes to address various resource allocation and network control problems ranging from congestion and power control to active queue management (AQM) in heterogeneous networks. Depending on the nature of the specific problem at hand, we utilize stochastic methods (Markov models), noncooperative game theory, hybrid systems, H-infinity optimal control, and pricing mechanisms. Active research directions in this area include rate allocation and flow assignment in heterogeneous wireless networks, spectrum sharing in cognitive radio settings, power control in optical networks, and novel stochastic approaches to AQM.
| Top of page |