Adaptive Orchestration for Proactive Security in Distributed Systems

This PhD position focuses on designing an adaptive orchestration framework for Moving Target Defense (MTD) in heterogeneous distributed systems. The research aims to integrate hardware-level insights, such as cache dynamics and performance counters, into the decision-making process of security orchestrators. By utilizing reinforcement learning, the candidate will develop agents capable of dynamically adjusting defense strategies based on real-time system context and vulnerability assessments. The project involves modeling service dependencies and evaluating the framework's effectiveness in terms of security, performance, and energy efficiency within an automotive use case.

• Define hardware-informed local security metrics focused on cache behavior.
• Develop graph-based models representing dependencies between services, resources, and attack surfaces.
• Design a unified reinforcement learning-based decision agent for adaptive MTD selection.
• Implement and evaluate the framework using a realistic automotive use case.
• Analyze trade-offs between security, system performance, and energy consumption.