In recent years machine learning has become increasingly mainstream across industries. Additionally, Graphical Processing Unit (GPU) accelerators are widely deployed in various neural network (NN) applications, including image recognition for autonomous vehicles and natural language processing, among others. Since training a powerful network requires expensive data collection and computing power, its design and parameters are often considered a secret intellectual property of their manufacturers. However, hardware accelerators can leak crucial information about the secret neural network designs through side-channels, like Electro-Magnetic (EM) emanations, power consumption, or timing. We propose and evaluate non-invasive and passive reverse engineering methods to recover NN designs deployed on GPUs through EM side-channel analysis. We employ a well-known technique of simple EM analysis and timing analysis of NN layers execution. We consider commonly used NN architectures, namely Multilayer Perceptron and Convolutional Neural Networks. We show how to recover the number of layers and neurons as well as the types of activation functions. Our experimental results are obtained on a setup that is as close as possible to a real-world device in order to properly assess the applicability and extendability of our methods. We analyze the NN execution of a PyTorch python framework implementation running on Nvidia Jetson Nano, a module computer embedding a Tegra X1 SoC that combines an ARM Cortex-A57 CPU and a 128-core GPU within a Maxwell architecture. Our results show the importance of side-channel protections for NN accelerators in real-world applications.

In profiling side-channel analysis, machine learning-based analysis nowadays offers the most powerful performance. This holds especially for techniques stemming from the neural network family: multilayer perceptron and convolutional neural networks. Convolutional neural networks are often favored as results suggest better performance, especially in scenarios where targets are protected with countermeasures. Multilayer perceptron receives significantly less attention, and researchers seem less interested in this method, narrowing the results in the literature to comparisons with convolutional neural networks. On the other hand, a multilayer perceptron has a much simpler structure, enabling easier hyperparameter tuning and, hopefully, contributing to the explainability of this neural network inner working. We investigate the behavior of a multilayer perceptron in the context of the side-channel analysis of AES. By exploring the sensitivity of multilayer perceptron hyperparameters over the attack’s performance, we aim to provide a better understanding of successful hyperparameters tuning and, ultimately, this algorithm’s performance. Our results show that MLP (with a proper hyperparameter tuning) can easily break implementations with a random delay or masking countermeasures. This work aims to reiterate the power of simpler neural network techniques in the profiled SCA.

Among the increasing evolution of IoT devices, practical applications need reliable secure protocols to communicate with each other. A major issue for modern cryptosystems is an implementation of secure and trustworthy mechanisms to rely on. A side-channel attack against these cryptosystems may overturn the guarantee of security against conventional cyber-attacks. Elliptic curve cryptography is public-key cryptography based on elliptic curves, and one of the well-known curves is Curve25519 which is used for TLS protocols as a recommended curve. This curve is mainly implemented on limited resource devices such as microcontrollers. However, this curve poses a weakness for low-order points during a Diffie-Hellman key exchange is employed. This research demonstrates possible exploitation of a threat of order 8 rational points of Curve25519 and shows results of the side-channel attacks using order 8 rational points on an embedded system. The results indicate the order 8 rational points might be applied to key extraction as attacker sides.

Profiling attacks, especially those based on machine learning proved as very successful techniques in recent years when considering side-channel analysis of block ciphers implementations. At the same time, the results for implementations of public-key cryptosystems are very sparse. In this paper, we consider several machine learning techniques in order to mount a power analysis attack on EdDSA using the curve Curve25519 as implemented in WolfSSL. The results show all considered techniques to be viable and powerful options. Especially convolutional neural networks (CNNs) are effective as we can break the implementation with only a single measurement in the attack phase while requiring less than 500 measurements in the training phase. Interestingly, that same convolutional neural network was recently shown to perform extremely well for attacking the implementation of the AES cipher. Our results show that some common grounds can be established when using deep learning for profiling attacks on distinct cryptographic algorithms and their corresponding implementations.