
We prove that the path-finding problem in isogeny graphs and the endomorphism ring problem for supersingular elliptic curves are equivalent under reductions of polynomial expected time, assuming the generalised Riemann hypothesis. The presumed hardness of these problems is foundational for isogeny-based cryptography. As an essential tool, we develop a rigorous algorithm for the quaternion analog of the path-finding problem, building upon the heuristic method of Kohel, Lauter, Petit and Tignol. This problem, and its (previously heuristic) resolution, are both a powerful cryptanalytic tool and a building-block for cryptosystems. This is an extended abstract of the full article available at http://arxiv.org/abs/2111.01481. This full article will be referred to as “the full version” throughout the text.
Since the beginning of the 21st century, modern information technology and electronic integrated circuit technology have developed rapidly. In the chip industry, the ability to resist side-channel attacks has become an important indicator for international mainstream evaluation agencies to evaluate chip security. This paper proposes an improved method for side channel analysis based on the $${CNN}_{best}$$ model, incorporating a lightweight combined channel and space convolutional attention module, optimising the position of the attention module, improving the learning efficiency of key features of the power consumption curve, and effectively reducing the number of traces used by the attack model. The addition of dropout layer network structure solves the problem that the model is prone to rapid overfitting. The optimal value of drop rate is sought through comparative experiments to speed up the convergence of the model and reduce the number of traces required for a successful attack. The experimental results show that the number of traces required by the method in this paper for side-channel attacks is reduced by 88% compared with the original model, which significantly improves the attack performance and can meet the requirements of side-channel modeling and analysis.
Federated learning has received extensive attention in recent years since the clients only need to share their local gradients with the servers without directly sharing their datasets to train the model. However, the existing research shows that the attackers can still reconstruct private information from shared gradients, resulting in privacy leakage. In addition, the aggregated results could be tampered with by servers or attackers. In this paper, we propose a secure and verifiable federated learning training scheme (SVFLS) to protect the privacy of data owners and verify aggregated results. Specifically, we employ threshold paillier encryption to protect the local gradients of data owners and use the bilinear aggregate signature to verify the correctness (or integrity) of aggregated results. Furthermore, our scheme can tolerate data owners dropping out during the training phase. We conduct extensive experiments on real datasets and demonstrate that our scheme is effective and practical.
Online restore reduces the downtime during backup restore, such that users can operate on the already restored files even if the other files are still being restored. However, due to the inconsistency between the access sequence and the restore sequence, the file to be accessed currently is sometimes not restored, thereby leading to long-delay time to access the file. We propose two approaches, which build on users’ historical access sequence to schedule the restore sequence, in order to reduce users’ waiting time to file access: (i) the frequency-based approach, which restores files based on the access frequencies of historical files; and (ii) the graph-based approach, which preferentially restores the frequently accessed files as well as their correlated files. Trace-driven experiments on two datasets show that our approaches significantly reduce users’ waiting time.
Port scanning attacks remain one of the major penetration testing schemes attackers employ to undertake maliferous intentions. With the increasingly sophisticated nature of cyber criminals and advanced technology and the failure of traditional network intrusion detection systems, the challenge of effectively detecting open ports with much efficiency in minimal time continues to linger. Thus, several recent studies, particularly those that employed machine learning approaches, have attempted to resolve and address the issue of enhancing this intrusion detection technique, yet suffer many performance challenges demanding further investigation. This paper employed seven machine learning classifiers to detect port scanning attacks after successfully using principal component analysis to resolve the relevant component and enhance the results. Comparison is made between the outcome of the various models and previous studies using accuracy, precision, recall, area-under-curve, f1-score, false-positive rate, and training time as performance metrics. Our results indicate that XGBoost was the best classifier with the highest accuracy of 99.98%, no false positive detected, a precision of 99.99%, a recall of 99.98, and an area-under-curve of 99.99% compared with the other classifiers and previous studies on port scan attack detection.