An Analog Side-Channel Attack on a High-Speed Asynchronous SAR ADC Using Dual Neural Network Technique

Abstract

SUMMARY This brief presents a side-channel attack (SCA) technique on a high-speed asynchronous successive approximation register (SAR) analog-to-digitalconverter(ADC).Theproposeddualneuralnetworkbasedonmultiplenoisewaveformsseparatelydisclosessignandabsolutevalue informationofinputsignalswhicharehiddenbythedifferentialstructureandhigh-speedasynchronousoperation.ThetargetSARADCandon-chip noisemonitorsaredesignedonasingleprototypechipforSCAdemon-stration.Fabricatedin40nm,theexperimentalresultsshowtheproposed attack on the asynchronous SAR ADC successfully restores the input data with a competitive accuracy within 300mV rms error.