Digital Signal Processing Reference
In-Depth Information
3
Secure Implementations: Side Channel Attacks
Cryptographic algorithms, i.e. mathematical objects describing the input-output
behavior of a black box, have to go through a thorough process of cryptanalysis
before they are widely accepted as secure and possibly standardized. Implementa-
tions of cryptographic algorithms essentially instantiate these black boxes.
An implementation of an algorithm does however not automatically inherit
the algorithm's security. The fact that the abstract black box is implemented in
software, hardware, or a combination of both gives rise to new security risks. Indeed,
measurable physical properties of an implementation become an additional, unin-
tended source of information, giving away knowledge about the secrets involved in
cryptographic implementations. Those sources are called side channels. While the
execution time of an implementation can be measured even from a distance, e.g.
over a network, the physical accessibility of embedded devices in particular gives
rise to even more side channels such as the power consumption, electromagnetic
radiation, acoustics, heat dissipation, light emission, etc.
Cryptanalytical methods exploiting such information leakage to extract secret
data are called side channel attacks. Side channel attacks are a serious concern as
they allow to extract secret information from unprotected implementations of black
box secure algorithms with moderate effort. With the ever increasing availability and
ubiquity of embedded devices, side channel attacks are a realistic threat nowadays.
One type of side channel attacks, a differential side channel attack, sequences
several steps to extract the secret information. In a first stage the attacker collects
measurements of the side channel, e.g. power consumption, as a function of the
time. Each measurement is the physical representation of the execution of the
cryptographic algorithm with a different message but a fixed key. In a second stage,
the attacker preprocesses the data, chooses a hypothetical power leakage model and
calculates the hypothetical leakage using this model for each message and for every
possible key guess. An example of one such model is the hamming weight of the
processed data at a certain time instant. In this case an adversary assumes that the
amplitude of the side channel measurement is related to the amount of binary ones in
last phase of the attack, the attacker uses a statistical test to quantify the similarity
between the hypothetical leakage and the real side channel leakage for every key
guess. The key that reveals the strongest relation is the adversary's best guess.
3.1
DSP Techniques Used in Side Channel Analysis
The side channel analysis research domain has become a mature research area
over the last decade, hence a multitude of techniques for preprocessing the data
and quantifying the relation between the real and hypothetical leakage have been
proposed.
