3 March 2025 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Human Activity Recognition (HAR) faces significant challenges when deployed in real-world scenarios due to non-independent and identically distributed (non-IID) data distributions. While existing domain adaptation (DA) approaches attempt to address this issue, they either require access to source data or struggle with large domain shifts. This paper presents a novel source-free domain adaptation framework for HAR that effectively handles substantial domain discrepancies across different datasets. Our approach introduces two key innovations: (1) a Discriminative Information Gramian (DIG) method that quantifies the relationship between target-domain samples and the source domain without requiring access to source data, and (2) an unsupervised domain generalization technique that ensures consistent feature extraction across augmented data samples, enhancing the model's effectiveness in the target domain. We evaluate our framework on five diverse HAR datasets comprising 87 users with varying demographics, devices, and environmental conditions. In single-source scenarios, our method achieves 76.77% accuracy and 67.03% F1-score, surpassing state-of-the-art solutions by 9.77% and 17.43%, respectively. For multi-source scenarios, we attain 85.33% accuracy and 79.55% F1-score, exceeding existing methods by at least 8.8% and 14.8%, respectively. This work represents the first successful attempt at dataset-level HAR domain adaptation without access to source data, marking a significant advancement in practical HAR applications.
1 March 2025 in IEEE Transactions on Dependable and Secure Computing
Backdoor attacks can exploit vulnerabilities in the training process of Deep Neural Networks (DNNs), introducing hidden malicious functionality that can be activated by a specific input pattern. Existing defenses typically rely on the assumption of a significant difference between poisoned and clean samples. However, subtle differences in dynamic and low poisoning ratio attacks can conceal poisoning features, thereby evading defenses. In this work, we propose a novel backdoor defense approach called ARTEMIS, which utilizes distribution shifts to eliminate the discrepancy between poisoned and benign samples in the feature space. Additionally, ARTEMIS learns from domain-invariant features after the shift. To further enhance the purification of backdoors in DNNs, we incorporate soft knowledge distillation into ARTEMIS to guide the alignment of features between the source dataset domain and the generated distribution-shift dataset domain. We extensively compare our proposed method with 5 state-of-the-art (SOTA) defensive techniques under 9 SOTA attacks on 4 datasets to demonstrate its effectiveness and robustness. Our results show that our method, ARTEMIS, can successfully purify dynamic and low poisoning ratio backdoor attacks and outperform existing defenses by a significant margin. We will release our code as open-source upon publication.