Profiled SCA with a New Twist: Semi-supervised Learning

Profiled side-channel attacks represent the most powerful category of side-channel attacks. In this context, the attacker gains access of a profiling device to build a precise model which is used to attack another device in the attacking phase. Mostly, it is assumed that the attacker has unlimited capabilities in the profiling phase, whereas the attacking phase is very restricted. We step away from this assumption and consider an attacker who is restricted in the profiling phase, while the attacking phase is less limited as in the traditional view. Clearly, in general, the attacker is not hindered to exchange any available knowledge between the profiling and attacking phase. Accordingly, we propose the concept of semi-supervised learning to side-channel analysis, in which the attacker uses the small amount of labeled measurements from the profiling phase as well as the unlabeled measurements from the attacking phase to build a more reliable model. Our results show that semi-supervised learning is beneficial in many scenarios and of particular interest when using template attack and its pooled version as side-channel attack techniques. Besides stating our results in varying scenarios, we discuss more general conclusions on semi-supervised learning for SCA that should help to transfer our observations to other settings in SCA.