CAPSL: The Component Authentication Process for Sandboxed Layouts

In this work, we propose a system-on-chip (SoC) design tool for the automatic generation of hardware sandboxes for securing untrusted IP to be integrated into trusted systems. The Component Authentication Process for Sandboxed Layouts (CAPSL) is a design flow that incorporates behavioral specifications of IP interfaces in order to generate sandboxes purposed for detecting trojan activation and isolating possible damage to a system at run-time. CAPSL adopts two formal models, interface automata and the Property Specification Language's sequential extended regular expressions (SERE), to generate reference monitors governing interactions of a collection of non-trusted IP. The sandbox partitions an untrusted sector that includes the non-secure IP and appropriate virtualized resources and controllers to isolate sandbox-system interactions upon deviation from the behavioral checkers. We review our design flow with an analysis of behavioral policy versatility and detection and defense mechanisms employed for various Trust-Hub.org benchmarks. Also presented is a brief resource evaluation highlighting CAPSL's reduced overhead compared to other run-time verification techniques.