Mathieu Renauld

Algebraic Side-Channel Attacks on the AES: Why Time also Matters in DPA

Algebraic side-channel attacks have been recently introduced as a powerful cryptanalysis technique against block ciphers. These attacks represent both a target algorithm and its physical information leakages as an overdefined system of equations that the adversary tries to solve. They were first applied to PRESENT because of its simple algebraic structure. In this paper, we investigate the extent to which they can be exploited against the AES Rijndael and discuss their practical specificities. We show experimentally that most of the intuitions that hold for PRESENT can also be observed for an unprotected implementation of Rijndael in an 8-bit controller. Namely, algebraic side-channel attacks can recover the AES master key with the observation of a single encrypted plaintext and they easily deal with unknown plaintexts/ciphertexts in this context. Because these attacks can take advantage of the physical information corresponding to all the cipher rounds, they imply that one cannot trade speed for code size (or gate count) without affecting the physical security of a leaking device. In other words, more intermediate computations inevitably leads to more exploitable leakages. We analyze the consequences of this observation on two different masking schemes and discuss its impact on other countermeasures. Our results exhibit that algebraic techniques lead to a new understanding of implementation weaknesses that is different than classical side-channel attacks.

Algebraic side-channel attacks

In 2002, algebraic attacks using overdefined systems of equations have been proposed as a potentially very powerful cryptanalysis technique against block ciphers. However, although a number of convincing experiments have been performed against certain reduced algorithms, it is not clear whether these attacks can be successfully applied in general and to a large class of ciphers. In this paper, we show that algebraic techniques can be combined with side-channel attacks in a very effective and natural fashion. As an illustration, we apply them to the block cipher PRESENT that is a stimulating first target, due to its simple algebraic structure. The proposed attacks have a number of interesting features: (1) they exploit the information leakages of all the cipher rounds, (2) in common implementation contexts (e.g. assuming a Hamming weight leakage model), they recover the block cipher keys after the observation of a single encryption, (3) these attacks can succeed in an unknown-plaintext/ciphertext adversarial scenario and (4) they directly defeat countermeasures such as boolean masking. Eventually, we argue that algebraic side-channel attacks can take advantage of any kind of physical leakage, leading to a new tradeoff between the robustness and informativeness of the side-channel information extraction.

A formal study of power variability issues and side-channel attacks for nanoscale devices

Variability is a central issue in deep submicron technologies, in which it becomes increasingly difficult to produce two chips with the same behavior. While the impact of variability is well understood from the microelectronic point of view, very few works investigated its significance for cryptographic implementations. This is an important concern as 65-nanometer and smaller technologies are soon going to equip an increasing number of security-enabled devices. Based on measurements performed on 20 prototype chips of an AES S-box, this paper provides the first comprehensive treatment of variability issues for side-channel attacks. We show that technology scaling implies important changes in terms of physical security. First, common leakage models (e.g. based on the Hamming weight of the manipulated data) are no longer valid as the size of transistors shrinks, even for standard CMOS circuits. This impacts both the evaluation of hardware countermeasures and formal works assuming that independent computations lead to independent leakage. Second, we discuss the consequences of variability for profiled side-channel attacks. We study the extend to which a leakage model that is carefully profiled for one device can lead to successful attacks against another device. We also define the perceived information to quantify this context, which generalizes the notion of mutual information with possibly degraded leakage models. Our results exhibit that existing side-channel attacks are not perfectly suited to this new context. They constitute an important step in better understanding the challenges raised by future technologies for the theory and practice of leakage resilient cryptography.

An Optimal key Enumeration Algorithm and Its Application to Side-Channel Attacks

Methods for enumerating cryptographic keys based on partial information obtained on key bytes are important tools in cryptanalysis. This paper discusses two contributions related to the practical application and algorithmic improvement of such tools. On the one hand, we observe that the evaluation of leaking devices is generally based on distinguishers with very limited computational cost, such as Kocher’s Differential Power Analysis. By contrast, classical cryptanalysis usually considers large computational costs (e.g. beyond 280 for present ciphers). Trying to bridge this gap, we show that allowing side-channel adversaries some computing power has major consequences for the security of leaking devices. For this purpose, we first propose a Bayesian extension of non-profiled side-channel attacks that allows us to rate key candidates according to their respective probabilities. Then we provide a new deterministic algorithm that allows us to optimally enumerate key candidates from any number of (possibly redundant) lists of any size, given that the subkey information is provided as probabilities, at the cost of limited (practically tractable) memory requirements. Finally, we investigate the impact of key enumeration taking advantage of this Bayesian formulation, and quantify the resulting reduction in the data complexity of various side-channel attacks.

Algebraic side-channel attacks beyond the hamming weight leakage model

Algebraic side-channel attacks (ASCA) are a method of cryptanalysis which allow performing key recoveries with very low data complexity. In an ASCA, the side-channel leaks of a device under test (DUT) are represented as a system of equations, and a machine solver is used to find a key which satisfies these equations. A primary limitation of the ASCA method is the way it tolerates errors. If the correct key is excluded from the system of equations due to noise in the measurements, the attack will fail. On the other hand, if the DUT is described in a more robust manner to better tolerate errors, the loss of information may make computation time intractable. In this paper, we first show how this robustness-information tradeoff can be simplified by using an optimizer, which exploits the probability data output by a side-channel decoder, instead of a standard SAT solver. For this purpose, we describe a way of representing the leak equations as vectors of aposteriori probabilities, enabling a natural integration of template attacks and ASCA. Next, we put forward the applicability of ASCA against devices which do not conform to simple leakage models (e.g. based on the Hamming weight of the manipulated data). We finally report on various experiments that illustrate the strengths and weaknesses of standard and optimizing solvers in various settings, hence demonstrating the versatility of ASCA.

Efficient removal of random delays from embedded software implementations using hidden markov models

Inserting random delays in cryptographic implementations is often used as a countermeasure against side-channel attacks. Most previous works on the topic focus on improving the statistical distribution of these delays. For example, efficient random delay generation algorithms have been proposed at CHES 2009/2010. These solutions increase security against attacks that solve the lack of synchronization between different leakage traces by integrating them. In this paper, we demonstrate that integration may not be the best tool to evaluate random delay insertions. For this purpose, we first describe different attacks exploiting pattern-recognition techniques and Hidden Markov Models. Using these tools and as a case study, we perform successful key recoveries against an implementation of the CHES 2009/2010 proposal in an Atmel microcontroller, with the same data complexity as against an unprotected implementation of the AES Rijndael. In other words, we completely cancel the countermeasure in this case. Next, we show that our cryptanalysis tools are remarkably robust to attack improved variants of the countermeasure, e.g. with additional noise or irregular dummy operations. We also exhibit that the attacks remain applicable in a non-profiled adversarial scenario. These results suggest that the use of random delays may not be effective for protecting small embedded devices against side-channel leakage. They highlight the strength of Viterbi decoding against such time-randomization countermeasures, in particular when combined with a precise description of the target implementations, using large lattices.

Fresh re-keying II: securing multiple parties against side-channel and fault attacks

Security-aware embedded systems are widespread nowadays and many applications, such as payment, pay-TV and automotive applications rely on them. These devices are usually very resource constrained but at the same time likely to operate in a hostile environment. Thus, the implementation of low-cost protection mechanisms against physical attacks is vital for their market relevance. An appealing choice, to counteract a large family of physical attacks with one mechanism, seem to be protocol-level countermeasures. At last years Africacrypt, a fresh re-keying scheme has been presented which combines the advantages of re-keying with those of classical countermeasures such as masking and hiding. The contribution of this paper is threefold: most importantly, the original fresh re-keying scheme was limited to one low-cost party (e.g. an RFID tag) in a two party communication scenario. In this paper we extend the scheme to n low-cost parties and show that the scheme is still secure. Second, one unanswered question in the original paper was the susceptibility of the scheme to algebraic SPA attacks. Therefore, we analyze this property of the scheme. Finally, we implemented the scheme on a common 8-bit microcontroller to show its efficiency in software.

Understanding the limitations and improving the relevance of SPICE simulations in side-channel security evaluations

Simulation is a very powerful tool for hardware designers. It generally allows the preliminary evaluation of a chip’s performance before its final tape out. As security against side-channel attacks is an increasingly important issue for cryptographic devices, simulation also becomes a desirable option for preliminary evaluation in this case. However, its relevance highly depends on the proper modeling of all the attack peculiarities. For example, several works in the literature directly exploit SPICE-like simulations without considering measurement peripherals. But the outcome of such analyses may be questionable, as witnessed by the recent results of Renauld et al. at CHES 2011, which showed how far the power traces of an AES S-box implemented using a dynamic and differential logic style fabricated in 65nm CMOS can lie from their post-layout simulations. One important difference was found in the linear dependencies between the (simulated and actual) traces and the S-box input/output bits. While simulations exhibited highly non-linear traces, actual measurements were much more linear. As linearity is a crucial parameter for the application of non-profiled side-channel attacks (which are only possible under the assumption of “sufficiently linear leakages”), this observation motivated us to study the reasons of such differences. Consequently, this work discusses the relevance of simulation in security evaluations, and highlights its dependency on the proper modeling of measurement setups. For this purpose, we present a generic approach to build an adequate model to represent measurement artifacts, based upon real data from equipment providers for our AES S-box case study. Next, we illustrate the transformation of simulated leakages, from highly non-linear to reasonably linear, exploiting our model and regression-based side-channel analysis. While improving the relevance of simulations in security evaluations, our results also raise doubts regarding the possibility to design dual-rail implementations with highly non-linear leakages.