Shadow is a block cipher for Internet of Things (IoT) Nodes proposed in the IEEE IoT Journal in 2021. The primary design principle of shadow is the adoption of a variant 4-branch Feistel structure to ensure a fast diffusion. We refer to this structure as the shadow structure and prove that it is almost identical to the Feistel structure, which invalidates the design principle. We also present a new structural distinguisher that can distinguish the shadow structure from a random permutation with only two plaintext/ciphertext pairs. Additionally, we demonstrate that the key-recovery attacks utilizing the impossible differential proposed by Liu et al. in the Cybersecurity Journal in 2023 and the integral characteristic proposed by Mirzaie et al. in the IEEE IoT Journal are infeasible. Instead, we extend our distinguisher to a key-recovery attack using only one plaintext/ciphertext pair by exploiting the key schedule. Moreover, upon investigating shadow’s round function, we observe that only specific forms of monomials can appear in the ciphertext, leading to an integral distinguisher involving four plaintext/ciphertext pairs. Notably, the algebraic degree does not exceed 12 for shadow-32 and 20 for shadow-64, regardless of the number of rounds used. Our results show that shadow is highly vulnerable to algebraic attacks, emphasizing the need for careful consideration of algebraic attacks when incorporating AND, rotation, and xor operations in cipher design.

A prominent countermeasure against side-channel attacks, the hiding countermeasure , typically involves shuffling operations using a permutation algorithm. This is especially crucial in the era of Post-quantum Cryptography, where computational characteristics of lattice and code-based cryptography heighten the need for robust defenses. In this context, securely and efficiently generating permutations is critical for an algorithm’s overall security and performance. Among the various approaches, the Fisher-Yates shuffle is widely adopted due to its security and ease of implementation. However, it is limited by a complexity of \(\mathcal {O}(N)\) due to its sequential nature. In response, we propose a time-area tradeoff swap algorithm, \(\mathsf {FSS}\) , that leverages a Butterfly Network structure, achieving only \(\log (N)\) depth, \(\log (N)\) work, and \(\mathcal {O}(1)\) operation time in parallel. Our analysis calculates the maximum gain an attacker can achieve through butterfly operations with \(\log (N)\) depth from a side-channel analysis perspective. Notably, we derive a generalized formula for the attack complexity of higher-order side-channel attacks for arbitrary input sizes, utilizing the fractal structure of the butterfly network. Moreover, our research demonstrates the efficiency and security of this permutation approach across different platforms. We include practical implementation results on ASIC as well as on CPU and GPU architectures, which underscore the algorithm’s performance advantages and robustness across diverse hardware environments. Through this exploration, we show that efficient and secure permutations can indeed be achieved with minimal randomness requirements.

Deep Neural Networks (DNNs) tend to perform poorly on unseen domains due to domain shifts. Domain Generalization (DG) aims to improve the performance on such scenarios by minimizing the distribution discrepancy between source domains. Among many studies, dropout-based DG approaches which remove domain-specific features have gained attention. However, they are limited in minimizing the upper bound of generalization risk because they do not explicitly consider the distribution discrepancy when discarding features. In this paper, we propose a novel Discrepancy-guided Channel Dropout (DgCD) for DG that explicitly derives the discrepancy between domains and drops the channels with significant distribution discrepancy. Given a training batch, we perform two ways of standardization: (1) based on the variance/mean of the batch (i.e., sampled from all source domains) and (2) based on the variance/mean of domain-wise samples in the batch. With the two normal distributions, we explicitly derive the discrepancy using KL-divergence and backpropagate it towards each channel. A channel with a higher contribution to the discrepancy is more likely to be dropped. Experimental results show the superiority of DgCD over the state-of-the-art DG baselines, demonstrating the effectiveness of our dropout strategy which is directly coupled to reducing the domain discrepancy. Our code is available at: https://github.com/gyeomo/DgCD

Model explainability is one of the crucial ingredients for building trustable AI systems, especially in the applications requiring reliability such as automated driving and diagnosis. Many explainability methods have been studied in the literature. Among many others, this article focuses on a research line that tries to visually explain a pre-trained image classification model such as Convolutional Neural Network by discovering concepts learned by the model, which is so-called the concept-based explanation. Previous concept-based explanation methods rely on the human definition of concepts (e.g., the Broden dataset) or semantic segmentation techniques like Slic (Simple Linear Iterative Clustering). However, we argue that the concepts identified by those methods may show image parts which are more in line with a human perspective or cropped by a segmentation method, rather than purely reflect a model's own perspective. We propose Model-Oriented Concept Extraction (MOCE), a novel approach to extracting key concepts based solely on a model itself, thereby being able to capture its unique perspectives which are not affected by any external factors. Experimental results on various pre-trained models confirmed the advantages of extracting concepts by truly representing the model's point of view.

Triangular meshes have been actively used in computer graphics to represent 3D shapes. However, due to their non-uniform and irregular nature, learning such data with a Deep Neural Network is not straightforward. Transforming mesh data to simpler structures (e.g., voxel grids, point clouds, or multi-view 2D images) leads to other issues including spatial information loss and scalability. Traditional descriptors for mesh data simply extract hand-crafted features, which might not be effective in various environments. Several deep architectures that directly consume mesh data have been proposed, but their input features are still heuristic and unable to fully capture both geodesic and geometric characteristics of a mesh. In addition, their model architectures are not designed to be capable of providing visual explanations of their decision making.