Mylène Roussellet

Horizontal correlation analysis on exponentiation

We introduce in this paper a technique in which we apply correlation analysis using only one execution power curve during an exponentiation to recover the whole secret exponent manipulated by the chip. As in the Big Mac attack from Walter, longer keys may facilitate this analysis and success will depend on the arithmetic coprocessor characteristics. We present the theory of the attack with some practical successful results on an embedded device and analyze the efficiency of classical countermeasures with respect to our attack. Our technique, which uses a single exponentiation curve, cannot be prevented by exponent blinding. Also, contrarily to the Big Mac attack, it applies even in the case of regular implementations such as the square and multiply always or the Montgomery ladder. We also point out that DSA and Diffie-Hellman exponentiations are no longer immune against CPA. Then we discuss the efficiency of known countermeasures, and we finally present some new ones.

Passive and Active Combined Attacks on AES Combining Fault Attacks and Side Channel Analysis

Tamper resistance of hardware products is currently a very popular subject for researchers in the security domain. Since the first Kocher side-channel (passive)attack, the Bellcore researchers and Biham and Shamir fault (active) attacks, many other side-channel and fault attacks have been published. The design of efficient countermeasures still remains a difficult task for IC designers and manufacturers as they must also consider the attacks which combine active and passive threats. It has been shown previously that combined attacks can defeat RSA implementations if side-channel countermeasures and fault protections are developed separately instead of being designed together. This paper demonstrates that combined attacks are also effective on symmetric cryptosystems and shows how they may jeopardize a supposedly state of the art secure AES implementation.

Improved collision-correlation power analysis on first order protected AES

The recent results presented by Moradi et al. on AES at CHES 2010 and Witteman et al. on square-and-multiply always RSA exponentiation at CT-RSA 2011 have shown that collision-correlation power analysis is able to recover the secret keys on embedded implementations. However, we noticed that the attack published last year by Moradi et al. is not efficient on correctly first-order protected implementations. We propose in this paper improvements on collision-correlation attacks which require less power traces than classical second-order power analysis techniques.We present here two new methods and show in practice their real efficiency on two first-order protected AES implementations. We also mention that other symmetric embedded algorithms can be targeted by our new techniques.

ROSETTA for Single Trace Analysis

In most efficient exponentiation implementations, recovering the secret exponent is equivalent to disclosing the sequence of squaring and multiplication operations. Some known attacks on the RSA exponentiation apply this strategy, but cannot be used against classical blinding countermeasures. In this paper, we propose new attacks distinguishing squaring from multiplications using a single side-channel trace. It makes our attacks more robust against blinding countermeasures than previous methods even if both exponent and message are randomized, whatever the quality and length of random masks. We demonstrate the efficiency of our new techniques using simulations in different noise configurations.

Simple power analysis on exponentiation revisited

Power Analysis has been studied since 1998 when P. Kocher et al. presented the first attack. From the initial Simple Power Analysis more complex techniques have been designed and studied during the previous decade such as Differential and Correlation Power Analysis. In this paper we revisit Simple Power Analysis which is at the heart of side channel techniques. We aim at showing its true efficiency when studied rigorously. Based on existing Chosen Message attacks we explain in this paper how particular message values can reveal the secret exponent manipulated during a modular exponentiation with a single power consumption curve. We detail the different ways to achieve this and then show that some blinded exponentiations can still be threatened by Simple Power Analysis depending on the implementation. Finally we will give advice on countermeasures to prevent such enhanced Simple Power Analysis techniques.

Side-Channel Analysis on Blinded Regular Scalar Multiplications

We present a new side-channel attack path threatening state-of-the-art protected implementations of elliptic curves embedded scalar multiplications. Regular algorithms such as the double-and-add-always and the Montgomery ladder are commonly used to protect the scalar multiplication from simple side-channel analysis. Combining such algorithms with scalar and/or point blinding countermeasures lead to scalar multiplications protected from all known attacks. Scalar randomization, which consists in adding a random multiple of the group order to the scalar value, is a popular countermeasure due to its efficiency. Amongst the several curves defined for usage in elliptic curves products, the most used are those standardized by the NIST. As observed in several previous publications, the modulus, hence the orders, of these curves are sparse, primarily for efficiency reasons. In this paper, we take advantage of this specificity to present new attack paths which combine vertical and horizontal side-channel attacks to recover the entire secret scalar in state-of-the-art protected elliptic curve implementations.

Square always exponentiation

Embedded exponentiation techniques have become a key concern for security and efficiency in hardware devices using public key cryptography. An exponentiation is basically a sequence of multiplications and squarings, but this sequence may reveal exponent bits to an attacker on an unprotected implementation. Although this subject has been covered for years, we present in this paper new exponentiation algorithms based on trading multiplications for squarings. Our method circumvents attacks aimed at distinguishing squarings from multiplications at a lower cost than previous techniques. Last but not least, we present new algorithms using two parallel squaring blocks which provide the fastest exponentiation to our knowledge.