Game Theory and Machine Learning for Cyber Deception, Resilience and Agility (GMC-DRA)
Abstract
Our objective is to support the development of new skills, with the aim of stimulating the development of security solutions that responds to problems that arise in various fields. We want to create a formal framework making it possible to contribute to the evolution of science and to the design and application of new techniques for solving security problems having consequences within the society. The main goal is described in two points: - An approach of cyber deception and network resilience to enhance networks security is to be developed. It leverages game-theoretic techniques to study the interactions between the attacker and the network defender. - A set up of a secure data exchange mechanism based on SDN (Software Defined Network) in the Cloud, with an impact on IoT infrastructures. These mechanisms must be able to face new threats and recover as soon as possible. Concerning the work statement and indicators, the tactical maneuvers attackers and defenders opt in the cyberspace bring us to divide the project in tasks. The main task being cyber deception. We are combining multiple deception techniques and network resilience leveraging the state-of-the-art in Machine Learning (ML) and Moving Target Defense (MTD). Our approach relies on using game theory and Reinforcement Learning (RL) to build models for new attack types. So, the following results from each task are expected: ¥ Design and Modeling of an Attack-Defense Epidemic Game. It will consist of developing novel epidemic model and characterize a partially observable stochastic game (POSG) model between the attacker and the defender. ¥ Develop Effective Strategies using Game Theory for Cyber Deception, Resilience, and Agility. It will be about to solve the developed POSG game model and propose a game solver to provide defense strategies and evaluate its effectiveness in terms of enhancing network resilience. It should also minimize the exploiting reasoning errors, the cognitive limitation and the concomitant biases that are the main factors making deception work. ¥ Enhance the Scalability, Adaptability, and Effectiveness of Cyber Deception and Resilience Game using Machine Learning. We will implement based game solving algorithm to solve large scale games and real-time security scenario. ¥ Develop a Moving Target Defense for Sustainable and Agile Security Posture in IoT Networks. This will consist of identify conditions for deception validity and reconfiguration periodicity; Then, design algorithm to implement MTD-based deception in IoT settings. ¥ Develop a Diversified Deception Algorithm for Secure Cloud Resilience. We will be able to integrate diversity-based deception in the developed strategies and identify optimal policy according to network settings. The impact of this project is to provide a complete algorithm to solve the associated security complex games using HSVI-based algorithm. The system should prove that the following metrics are efficient: System Impact, Total recovery Effort, Recovery Dependent resilience.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 07, 2021
- Source ID
- W911NF2110326
Entities
People
- Marcellin Nkenlifack
Organizations
- Army Contracting Command
- United States Army
- Université de Dschang