Mind the Portability: A Warriors Guide through Realistic Profiled Side-channel Analysis

Abstract

Profiled side-channel attacks represent a practical threat to digital devices, thereby having the potential to disrupt the foundation of e-commerce, the Internet of Things (IoT), and smart cities. In the profiled side-channel attack, the adversary gains knowledge about the target device by getting access to a cloned device. Though these two devices are different in realworld scenarios, yet, unfortunately, a large part of research works simplifies the setting by using only a single device for both profiling and attacking. There, the portability issue is conveniently ignored to ease the experimental procedure. In parallel to the above developments, machine learning techniques are used in recent literature, demonstrating excellent performance in profiled side-channel attacks. Again, unfortunately, the portability is neglected. In this paper, we consider realistic side-channel scenarios and commonly used machine learning techniques to evaluate the influence of portability on the efficacy of an attack. Our experimental results show that portability plays an important role and should not be disregarded as it contributes to a significant overestimate of the attack efficiency, which can easily be an order of magnitude size. After establishing the importance of portability, we propose a new model called the Multiple Device Model (MDM) that formally incorporates the device to device variation during a profiled side-channel attack. We show through experimental studies how machine learning and MDM significantly enhance the capacity for practical side-channel attacks. More precisely, we demonstrate how MDM can improve the performance of an attack by order of magnitude, completely negating the influence of portability.

A Systematic Side-Channel Evaluation of Black Box AES in Secure MCU: Architecture Recovery and Retrieval of PUF Based Secret Key

“…Figure 5 shows the result (only two key bytes are shown for brevity) for attack in self and cross device setting. Indeed, there is degradation of 30-60% degradation in number of traces required, however this is much better than 2000% degradation reported for other devices [13]. 60% degradation correspond to few extra minutes of measurement time, making cross-device a serious threat to such hardware engines.…”

“…only a few traces from victim device and leakage profile from the clone device, performs a key recovery attack on victim device. The use of deep neural networks (DNN) have made cross-device attacks effective [13], [14]. DNN are trained with SCA measurements from clone device with known leakage labels, while trained model is queried with unlabeled SCA traces from victim device.…”

“…In cross-device setting, the clone device and victim device might differ due to process variations, making the learned model overfit for clone device. For example, [13] reported up to 20× degradation in attack performance in cross-device setting. We measured traces from two identical OKdo E1 development board: 100, 000 traces are measured from Device 1 for training the DNN while 20, 000 traces from Device 2 (victim).…”

Online Template Attack on ECDSA:

“…The work of Choudary and Kuhn focuses on differences between devices when performing portable template attacks (considering mainly DC offset caused by temperature changes) [7]. Bhasin et al recently compared different machine learning techniques on the efficacy with respect to portability [3]. A common point in all those works is they identified the problem and try to amortize the penalty due to portability.…”

“…However, profiling traces obtained from a device different than the target device are not necessarily accurate due to variations between individual chips. This is the so-called issue of portability that has received wide coverage in many papers the last few years [3,6]. For dedicated security products such as smart cards it is often even infeasible to obtain an open device similar to the target device and profiling is simply out of the question.…”