Federated Recommender Systems (FedRecs) leverage federated learning to protect user privacy by retaining data locally. However, user embeddings in FedRecs often encode sensitive attribute information, rendering them vulnerable to attribute inference attacks. Attribute unlearning has emerged as a promising approach to mitigate this issue. In this paper, we focus on user-level FedRecs, which is a more practical yet challenging setting compared to group-level FedRecs. Adversarial training emerges as the most feasible approach within this context. We identify two key challenges in implementing adversarial training-based attribute unlearning for user-level FedRecs: i) mitigating training instability caused by user data heterogeneity, and ii) preventing attribute information leakage through gradients. To address these challenges, we propose FedAU2, an attribute unlearning method for user-level FedRecs. For CH1, we propose a adaptive adversarial training strategy, where the training dynamics are adjusted in response to local optimization behavior. For CH2, we propose a dual-stochastic variational autoencoder to perturb the adversarial model, effectively preventing gradient-based information leakage. Extensive experiments on three real-world datasets demonstrate that our proposed FedAU2 achieves superior performance in unlearning effectiveness and recommendation performance compared to existing baselines.

Backdoor attacks represent a significant threat to the security of deep learning models. Due to the stealthiness of backdoor attacks, effectively detecting whether a model has been compromised by such attacks remains a major challenge. Previous backdoor detection methods either rely on backdoor datasets to identify anomalies in backdoor samples or depend on prior knowledge of existing backdoor attacks. This results in difficulties in detecting backdoors when backdoor samples are unavailable and leads to poor generalization capabilities when addressing new attack methods. This work proposes a novel approach for detecting backdoor attacks called Distill To Detect (D2D), that does not depend on backdoor samples or any prior knowledge. It utilizes knowledge distillation to amplify more general and universal backdoor anomalies exhibited on clean samples for detection. This approach is not only more efficient but also exhibits strong generalization capabilities, enabling the detection of most backdoor attacks with low time and computational costs. We tested our approach against six backdoor attacks and three different model architectures, demonstrating the effectiveness of our proposed method.

ABSTRACT In the logit‐based knowledge distillation method, the student model learns the classification information of the teacher network by transmitting high‐dimensional and abstract logits. Nevertheless, the teacher network is not an optimal learning target. On common datasets such as CIFAR100 and ImageNet, the majority of models exhibit classification accuracies of only 60% to 80%. These errors in the teacher models are a significant part of knowledge distillation that cannot be ignored. In order to facilitate the acquisition of more accurate knowledge by students, we propose the implementation of adaptive logit reconstruction knowledge distillation (ALRKD). ALRKD corrects errors by using the standard deviation, which represents the fluctuation degree of the logit distribution. Furthermore, in order to compensate for the loss of information that occurs during the correction process, an additional branch is designed to provide supplementary knowledge regarding the relationships between other classes. The results of several experiments on common datasets demonstrate the significant superiority of ALRKD.

As a variant of the classical knapsack problem, the multi-dimension multi-choice knapsack problem (MMKP) is widely used in practical applications. It is a NP-complete problem, the exact solution of MMKP cannot be founded in polynomial-time. As one of the heuristic algorithms, quantum particle swarm optimization (QPSO) algorithm provides a sight to get the approximately optimal result for MMKP. However, due to the multiple constraints among dimensions and the dispersing feasible regions, QPSO tends to fall into local convergence. Hence a modified diversity reserved QPSO algorithm for MMKP is proposed in this paper: (i) to measure the availability of a particle by comparing the position between itself and the next alternative during the generation; (ii) import a position disturbance operator to increase the diversity of population. Experiments demonstrate that the proposed evolutionary algorithm could find better near-optimal results. And the analysis of convergence and execution time suggest that the probability of local convergence is declined in our algorithm.

Feature selection is an important part of machine learning and data mining which may enhance the speed and the performance of learning and mining algorithms. Given certain criteria to evaluate features, the problem of feature selection can be regarded as an optimization problem. Therefore, evolutionary algorithms can be used to solve such a kind of optimization problems. In this paper, we present a novel feature subset selection approach based on the framework of genetic algorithms. Two new mutation operators are constructed using the standard deviation of candidate features and the cardinality of candidate feature subsets. Then, a filter feature subset selection approach using a dynamic mixed strategy is proposed, which combines the new mutation operators with the single-point mutation operator. The new approach can not only dynamically adjust the probability distribution over these three mutation operators, but also maintain the combined effects of feature subsets as a whole fitness evaluation. The proposed approach is able to quickly escape from local optimal feature subsets and to obtain smaller scale subsets than evolutionary algorithms using a single mutation operator. Experiments have been implemented on six standard UCI datasets and the proposed algorithm is compared with other classical algorithms. The comparison outcomes confirm the effectiveness of our approach.