Power fingerprinting in SDR integrity assessment for security and regulatory compliance

Software-Defined Radio (SDR) provides a flexible platform that facilitates radio resource management and enables new technologies and applications. Unfortunately, their reliance on software implementations makes them vulnerable to malicious software attacks that could impact their spectral emissions and disclose sensitive information. It is of critical importance for the widespread deployment of SDR to develop technologies that enable effective integrity assessment of communications platforms and timely detection of malicious intrusions. We provide further evidence of the feasibility of a novel approach called Power Fingerprinting (PFP) that enables an effective mechanism to perform integrity assessment of SDR. PFP relies on an external monitor that captures fine-grained measurements of the processor's power consumption and compares them against stored signatures from trusted software by applying pattern recognition and signal detection techniques. Because it is implemented by an external monitor, PFP causes minimal disruption on the target system and also provides the necessary isolation to protect against malicious attacks to the monitor itself. Fine-granularity measurements deliver improved visibility into the execution status and make the PFP monitor difficult to evade, while the reliance on anomaly detection from trusted references makes it effective against zero-day attacks. We present the results of different feasibility experiments that support the applicability of PFP to SDR integrity assessment. In the first experiment, a PFP monitor is able to effectively detect the execution of a tampered routine that misconfigures the operational mode of a PICDEM Z radio platform, affecting the resulting spectral emission. In a second experiment, our monitor effectively identifies when a transmission routine is modified, affecting encryption settings. We also present an approach to improve the performance of PFP by characterizing the way a specific platform consumes power. This platform characterization, which can be done using principal component analysis or linear discriminant analysis, allows a PFP monitor to work only on the features that carry the most information. As a result, the PFP monitor is able to detect execution deviations resulting from a difference of a single bit transition, the smallest possible disruption.