Research Seminar - Utilization of Machine Learning Algorithms in Securing Endpoints of Communication: Ensemble Trust Management

Dr. Khalil Abuosba, a candidate for the Orillia LTA position in the Department of Computer Science, will present a research seminar.

Title: Utilization of Machine Learning Algorithms in Securing Endpoints of Communication: Ensemble Trust Management Methods

Abstract: Operational computing environment is viewed as a finite state machine that may be specified based on the operational requirements. Reduction of exposure within an operational computing environment leads to reduction of operational risk; hence, more- reliable and survivable computing environments would be achieved.

Computational models have been transformed into new concepts that have distributed nature; whether it is web-based services or IoT solutions; securing points of communications is one of the most challenging requirements of deployment. In this seminar we focus on distributed systems trust management methodologies that are efficient and reliable. Traditional security methods are strictly based on implement security controls and enforce security policies that would satisfy the security requirements for the purpose of resolving known threats that might compromise the endpoint of communication.

This approach lacks protection against anomalous threats. A risk management approach to resolve anomaly is trust management. In this research, We explore deployment of rational agents that gather information and learn from the precepts they it receive from the environment, whether partially observable or fully observable.

We also map agent characteristics to the environment; IoT solutions mostly are deterministic in nature; however, web transactions are nondeterministic or stochastic in nature. Security trust management problems are resolved by satisfying constraints. Another dimension that we address is resolving issues of uncertainty. Finally, we will be addressing the learning algorithms that would be suitable for trust management including decision trees, ensemble methods, and reinforcement learning. We also highlight on neural networks as an application for decision making. Formal methods as well as complexity issues are addressed as well.

Zoom link

https://lakeheadu.zoom.us/j/93172465169?pwd=M0psbks3L1BLNkRhY01pSG9ucktD... Meeting ID: 931 7246 5169 Passcode: 6448