There have been various attempts to attack reduced variants of Trivium stream cipher using cube attack. During the preprocessing phase of cube attack, we need to test the linearity of a superpoly. The linearity testing problem is to check whether a function is close to linear by asking oracle queries to the function. This is the BLR linearity test for Boolean functions, which has a time complexity of O(22k + c) cipher operations, where k is the length of the key and c is the size of the cube. In this paper we present a method which is supposed to be a sufficient condition for testing a superpoly for linearity in F2 with a time complexity O(2c + 1 (k2 + k)). Our analysis on Trivium cipher reduced to 576 rounds using cube attack gives 69 extremely sparse linearly independent linear equations for smaller cubes, which recovers 69 bits of the key and reduces the attack complexity in the online phase to 211.
This paper provides an in-depth exploration of neural distinguishers in cryptanalysis with a focus on the SLIM Cipher. Building upon the work of Gohr, we aim to explain why neural distinguishers outperform traditional differential attacks and enhance contemporary methodologies. Our experiments centered on SLIM Cipher's fifth round and aimed to understand the influence of input difference, output difference, and previous round differentials on the accuracy of the neural distinguisher. Our investigations revealed that the choice of input difference was significant, and only focusing on maximizing the differential probability might not always yield the best results. Neural distinguisher appears to observe not just about the distribution of differences from the ciphertext but also additional cryptanalytic features or properties, which was a key insight from our research. Furthermore, we found that the output difference in the ciphertext played a vital role in the neural distinguisher's performance. By partially decrypting ciphertext pairs, we discovered that the output difference from the penultimate round was also a major factor impacting the accuracy of the neural distinguisher. This led to an improved methodology, yielding impressive accuracy results. These findings present valuable insights that can guide the future development and application of these tools in cryptanalysis.
This study presents a robust version of Light Cipher Block (LCB) by addressing the vulnerabilities identified in previous versions. The vulnerabilities in LCB, including a linear S-Box, improper bit shuffling, and subkey reusability, were thoroughly examined. To overcome these weaknesses, a modified version called Secure LCB is introduced, incorporating changes to the P-Box and key generation algorithm. Motivated by Gohr’s work at CRYPTO’19, this paper investigates the use of a neural distinguisher built upon a 1-dimensional convolutional neural network (1-d CNN). The deep learning model is tasked with identifying ciphertexts that have a specific, controlled difference in their inputs, as opposed to those with random input differences. The evaluation of the proposed Secure LCB using the neural distinguisher suggests that the modifications made to LCB have effectively enhanced its resistance against the neural distinguisher’s classification. This highlights the importance of addressing vulnerabilities in cryptographic systems and showcases the potential of machine learning techniques in cryptanalysis.