GNN-Based Reverse Engineering Protection Using Obfuscation Through Electromagnetic Interference

Abstract

To combat PCB design reverse engineering and protect intellectual property, a method of obfuscation is developed. Noise generating traces are placed alongside the true circuit, separated by false non-conductive vias visually indistinguishable from regular ones. If the false vias are copied as real conductive vias in illegitimate replications, signals will be distorted, leading to poor performance. The deliberate placement of the generators separated by false vias is tested using a machine learning algorithm (MLA) employing a signal integrity database to determine the most vulnerable traces in the PCB based on their physical properties and interconnections.