Predicting Secret Keys Via Branch Prediction

Abstract

Abstract. This paper announces a new software side-channel attackenabled by the branch prediction capability common to all modern highperformance CPUs. The penalty paid (extra clock cycles) for a mispredicted branch can be used for cryptanalysis of cryptographic primitives that employ a data-dependent program flow. Analogous to the recently described cache-based side-channel attacks our attacks also allow an unprivileged process to attack other processes running in parallel on the same processor, despite sophisticated partitioning methods such as memory protection, sandboxing or even virtualization. In this paper, we will discuss several such attacks for the example of RSA, and experimentally show their applicability to real systems, such as OpenSSL and Linux. Moreover, we will also demonstrate the strength of the branch prediction sidechannel attack by rendering the obvious countermeasure in this context (Montgomery Multiplication with dummy-reduction) as useless. Although the deeper consequences of the latter result make the task of writing an efficient and secure modular exponentiation (or scalar multiplication on an elliptic curve) a challenging task, we will eventually suggest some countermeasures to mitigate branch prediction side-channel attacks.

An overview of cryptanalysis research for the advanced encryption standard

“…Another method exploits timing dependencies between the branch prediction capability common to all high performance micros and bits of the secret key [33], [34]. Compromise of the RSA algorithm was successfully demonstrated using this attack and generalization to symmetric ciphers is recommended for future work in [35]. These attacks are also applicable to hardware implementations that utilize these same processing elements.…”

Manger’s Attack Revisited

“…As a consequence, it cannot be excluded that the conditional branching can be detected through the timing in certain scenarios. Furthermore, based on the conditional branching, the attack can in principle be mounted as a branch prediction attack [7]. Basically, it is easily possible to remove this vulnerability by using branch free code employing the techniques shown in [8,9].…”

Securing Openssl Against Micro-Architectural Attacks

“…More sophisticated attacks are presented in (Acıiçmez et al, 2007b;Acıiçmez et al, 2007c) that modify the BTB to produce effects that can leak information more efficiently than observing the time taken to compute an algorithm. Indeed, the most efficient attack described involves closely observing the BP during the computation of an RSA signature by using a spy process that modifies the BTB and observes the subsequent behaviour.…”

“…This could allow an attacker to derive the private key from one signature generation. An implementation of this type of attack on a modified version of the function used in OpenSSL to generate RSA signatures is described in (Acıiçmez et al, 2007b).…”