Differential Power Analysis using wavelet decomposition

Differential Power Analysis (DPA) has been successfully used against crytographic hardware to extract the secret key in a non-invasive manner. However, the great success of DPA requires a large number of traces to overcome system noise or countermeasures, which equates to increased processing times and computing hardware requirements. We investigate wavelet decomposition of a DPA trace data set as a means to reducing the number of traces. By decomposing the signal into various wavelet coefficient levels, we identify those that reduce DPA performance and mitigate their impact. We achieve an 11.53% increase in correct key correlation value vs traditional DPA and exceed traditional DPA with as little as 30 traces. This method significantly reduces the number of traces needed to overcome system noise and counter-measures which introduce random operations.