Descrição
Side-Channel Analysis (SCA) represents a serious threat to the security of millions of smart devices that form part of the so-called Internet of Things (IoT). On the other hand, perform the "right- fitting" cryptographic code for the IoT is a highly challenging task due to the reduced resource constraints of must of the IoT devices and the variety of cryptographic algorithms on disposal. An important criterion to assess the suitability of a light-weight cipher implementation, with respect to the SCA point of view, is the amount of energy leakage available to an adversary. In this thesis, the efficiency of a selected function that is commonly used in AES implementations in the perspective of Correlation Power Analysis (CPA) attacks are analyzed, leading to focus on the very common situation where the exact time of the sensitive processing is drowned in a large number of leakage points. In the particular case of statistical attacks, much of the existing literature essentially develop the theory under the assumption that the exact sensitive time is known and cannot be directly applied when the latter assumption is relaxed, being such a particular aspect for the simple Differential Power Analysis (DPA) in contrast with the CPA. To deal with this issue, an improvement that makes the statistical attack a real alternative compared with the simple DPA has been proposed. For the power consumption model (Hamming Weight model), and by rewriting the simple DPA attacks in terms of correlation coefficients between Boolean functions. Exhibiting properties of S-boxes relied on CPA attacks and showing that these properties are opposite to the non-linearity criterion and to the propagation criterion assumed for the former DPA. In order to achieve this goal, the study has been illustrated by various attack experiments performed on several copies implementations of the light-weight AES chipper in a well-known micro-controller educative platform within an 8-bit processor architecture deployed on a 350 nanometers CMOS technology. The Side-channel attacks presented in this work have been set in ideal conditions to capture the full complexity of an attack performed in real-world conditions, showing that certain implementation aspects can influence the leakage levels. On the other side, practical improvements are proposed for specific contexts by exploring the relationship between the non-linearity of the studied selection function and the measured leakages, with the only pretension to bridge the gap between the theory and the practice. The results point to new enlightenment on the resilience of basic operations executed by common light-weight ciphers implementations against CPA attacks.
