A game theory approach for RTL security verification resources allocation

Abstract

Many Trojan detection technologies are too time-consuming to cover the entire state space in complex designs. The valuable verification resources should be allocated to regions vulnerable to security threats. However, there are few studies on security verification resources allocation. To fill in this gap, we design a security game framework to guide the security verification resources allocation. The framework utilizes the Trojan vulnerability measurement as player utilities, so the utility value determination doesn't need any expert prior knowledge to the specific design under test. A new Stackelberg security game specific to hardware security is also proposed. The new game model minimizes the defender utility loss with the limited verification resources restriction. Due to the lack of study on RTL Trojan vulnerability measurement, we also propose a RTL security vulnerability measurement to measure each logic propagation path vulnerability quantitatively and efficiently. We apply the proposed Stackelberg security game framework to Trust-hub Trojan benchmarks written by Verilog RTL code. The experiments demonstrate that the most suspicious logic propagation path is one part of Trojan in most cases and the proposed RTL security vulnerability measurement is effective. Also, the allocation strategy calculated by security game could get security confidence as high as possible with all available resources and may also cover the Trojan even when the carefully design Trojan evade the vulnerability measurement.