Enhancing Power Side-Channel Detection of Attacks to Additive Manufacturing Systems

Abstract

Additive Manufacturing (AM) allows the production of system-critical parts in distributed manufacturing environments. This just-in-time process is becoming increasingly used for life, health, or mission-critical parts. AM allows these parts to be manufactured rapidly with complex and internal geometries. Attackers can remotely and covertly cause anomalies in these parts, which include internal voids or mechanical weaknesses that are not detectable using conventional inspection methods. In this work, we enhance the power side-channel analysis of AM systems, which has been shown to detect many sabotage attacks by measuring the current delivered to the AM machine's actuators throughout a print. Time-frequency analysis is performed on the collected power data and compared to the expected power signature to detect attacks. We enhance this process by identifying key algorithm parameters, their effect on attack detectability, and their effect on required computational resources. By manipulating these parameters, it is possible to increase the detectability of attacks. Through this, we demonstated the detction of extrusion rate change attacks and smart void attacks; both previously thought undetectable with this method. The parameters that have the most significant impact on detectability and computation time have been identified, and recommendations are given to increase detectability across the attack surface while minimizing time for analysis. This thesis identifies, analyzes, and utilizes key alogithm parameters for power side channel attack detection in additive manufacturing systems.