Abstract:This paper proposes and evaluates the applications of different DL techniques including the Convolutional Neural Network (CNN) and the multilayer perceptron (MLP) models for non‐profiled attacks on the AES‐128 encryption implementation. Along with the designed models, a dataset reconstruction and labelling technique for the proposed model has also been performed for solving the high dimension data and imbalanced dataset problem. As a result, the DL‐based SCA with our reconstructed dataset for different targets of ASCAD, RISC‐V microcontroller, and ChipWhisperer boards has achieved a higher performance of non‐profiled attacks. The experimental results have clarified that the exponential linear unit (ELU) function is better than the rectified linear unit (ReLU) in taking the correct key with the presence of the Gaussian noise. In modern embedded systems, security issues including side‐channel attacks (SCAs) are becoming of paramount importance since the embedded devices are ubiquitous in many categories of consumer electronics. Recently, deep learning (DL) has been introduced as a new promising approach for profiled and non‐profiled SCAs. This paper proposes and evaluates the applications of different DL techniques including the Convolutional Neural Network and the multilayer perceptron models for non‐profiled attacks on the AES‐128 encryption implementation. Especially, the proposed network is fine‐tuned with different number of hidden layers, labelling techniques and activation functions. Along with the designed models, a dataset reconstruction and labelling technique for the proposed model has also been performed for solving the high dimension data and imbalanced dataset problem. As a result, the DL based SCA with our reconstructed dataset for different targets of ASCAD, RISC‐V microcontroller, and ChipWhisperer boards has achieved a higher performance of non‐profiled attacks. Specifically, necessary investigations to evaluate the efficiency of the proposed techniques against different SCA countermeasures, such as masking and hiding, have been performed. In addition, the effect of the activation function on the proposed DL models was investigated. The experimental results have clarified that the exponential linear unit function is better than the rectified linear unit in fighting against noise generation‐based hiding countermeasure.

How Diversity Affects Deep-Learning Side-Channel Attacks

Deep Learning Side-Channel Collision Attack

Deep learning side-channel attack against hardware implementations of AES

Research on Side-Channel Analysis Based on Deep Learning with Different Sample Data

Adversarial Attack Based Countermeasures against Deep Learning Side-Channel Attacks

Power side-channel leakage localization through adversarial training of deep neural networks

A Deep-Learning Technique to Locate Cryptographic Operations in Side-Channel Traces

DeepTheft: Stealing DNN Model Architectures through Power Side Channel

Practical Approaches Towards Deep-Learning Based Cross-Device Power Side Channel Attack

Generalized Power Attacks against Crypto Hardware using Long-Range Deep Learning

Stealthy and Robust Glitch Injection Attack on Deep Learning Accelerator for Target with Variational Viewpoint.

Leaky DNN: Stealing Deep-Learning Model Secret with GPU Context-Switching Side-Channel

Practical Approaches Toward Deep-Learning-Based Cross-Device Power Side-Channel Attack

Practical aspects on non-profiled deep-learning side-channel attacks against AES software implementation with two types of masking countermeasures including RSM

On the performance of non‐profiled side channel attacks based on deep learning techniques

Design of Convolutional Neural Networks Architecture for Non-Profiled Side-Channel Attack Detection

Investigating Efficient Deep Learning Architectures For Side-Channel Attacks on AES

Deep learning‐assisted and combined attack: a novel side‐channel attack

I Know What You See: Power Side-Channel Attack on Convolutional Neural Network Accelerators

Acnn: arbitrary trace attacks based on leakage area detection

Amplitude-modulated EM side-channel attack on provably secure masked AES