Physical layer identification of embedded devices using RF-DNA fingerprinting

Abstract

RF distinct native attribute (RF-DNA) fingerprinting is introduced as a means to uniquely identify embedded processors and other integrated circuit devices by passively monitoring and exploiting unintentional RF emissions. Device discrimination is accomplished using RF-DNA fingerprints comprised of higher-order statistical features based on instantaneous amplitude and frequency responses as a device executes a sequence of operations. The resultant fingerprints are input to a Multiple Discriminant Analysis/Maximum Likelihood (MDA/ML) processor for subsequent device discrimination. Using devices from a given manufacturer and experimentally collected side channel signals, 90-100% identification accuracy is achieved for SNR ≥ 12 dB for devices with identical part numbers from the same production lot. Depending on the level of required classification accuracy, RF-DNA fingerprinting is well-suited for realistic environments and practical operating distances. Applications of device RF-DNA fingerprints include supplementary physical layer authentication of secure tokens (e.g. smart cards), detection of counterfeit electronic devices or unauthorized modification, and forensic attribution of a device's unique identity in criminal or other investigations.