Side-channel attacks represent one of the most powerful
category
of attacks on cryptographic devices with profiled attacks in a
prominent place as the most powerful among them. Indeed, for instance,
template attack is a well-known real-world attack that is also the most
powerful attack from the information theoretic perspective. On the other
hand, machine learning techniques have proven their quality in a numerous
applications where one is definitely side-channel analysis, but they
come with a price. Selecting the appropriate algorithm as well as the
parameters can sometimes be a difficult and time consuming task.
Nevertheless,
the results obtained until now justify such an effort.
However, a large part of those results use simplification of the data
relation from the one perspective and extremely powerful machine
learning techniques from the other side. In this paper, we concentrate
first on the tuning part, which we show to be of extreme importance.
Furthermore, since tuning represents a task that is time demanding, we
discuss how to use hyperheuristics to obtain good results in a relatively
short amount of time. Next, we provide an extensive comparison between
various machine
learning techniques spanning from extremely simple
ones ( even without any parameters to tune), up to methods where
previous experience
is a must if one wants to obtain competitive
results. To support our claims, we give extensive experimental results
and discuss the necessary
conditions to conduct a proper machine
learning analysis. Besides the machine learning algorithms' results, we
give results obtained with the template attack. Finally, we investigate the
influence of the feature (in)dependence in datasets with varying amount
of noise as well as the influence of feature noise and classification noise. In
order to strengthen our findings, we also discuss provable machine
learning algorithms, i.e., PAC learning algorithms.
