Legacy software written in unsafe languages is inherently error-prone and offers the model breeding ground for security vulnerabilities that are exploited in the wild. A key solution to deal with this challenge is to compartmentalize software such that different compartments run with the least set of privileges required to perform the underlying task. While compartmentalization can effectively limit the damage due to an initial penetration, enforcing such strategies at a fine granularity, while providing high performance, efficiency, and scalability is a critical challenge that this proposal seeks to solve.

The goal of the HERCULES project is to provide a holistic compartmentalization solution, following the principle of least privilege, that spans multiple layers of the computing stack. The project integrates several novel solutions, including sound and precise static analysis methods, dynamic tracing with high coverage, principal modeling, effective source code annotations, automated code restructuring, and low-level compiler transformations. HERCULES has three key thrusts:

• static analysis for automated inference of global, principal-agnostic access control policies,
• dynamic analysis for automated inference of principal-specific access control policies, and
• code restructuring for automatic creation of fine-grained compartments.