Jiaping Yu

Jiaping Yu, Muli Yang, Jiapeng Ji et al.

Source-Free Unsupervised Domain Adaptation (SFUDA) addresses the realistic challenge of adapting a source-trained model to a target domain without access to the source data, driven by concerns over privacy and cost. Existing SFUDA methods either exploit only the source model's predictions or fine-tune large multimodal models, yet both neglect complementary insights and the latent structure of target data. In this paper, we propose the Experts Cooperative Learning (EXCL). EXCL contains the Dual Experts framework and Retrieval-Augmentation-Interaction optimization pipeline. The Dual Experts framework places a frozen source-domain model (augmented with Conv-Adapter) and a pretrained vision-language model (with a trainable text prompt) on equal footing to mine consensus knowledge from unlabeled target samples. To effectively train these plug-in modules under purely unsupervised conditions, we introduce Retrieval-Augmented-Interaction(RAIN), a three-stage pipeline that (1) collaboratively retrieves pseudo-source and complex target samples, (2) separately fine-tunes each expert on its respective sample set, and (3) enforces learning object consistency via a shared learning result. Extensive experiments on four benchmark datasets demonstrate that our approach matches state-of-the-art performance.

Haotian Chi, Qiang Zeng, Xiaojiang Du et al.

A number of Internet of Things (IoTs) platforms have emerged to enable various IoT apps developed by third-party developers to automate smart homes. Prior research mostly concerns the overprivilege problem in the permission model. Our work, however, reveals that even IoT apps that follow the principle of least privilege, when they interplay, can cause unique types of threats, named Cross-App Interference (CAI) threats. We describe and categorize the new threats, showing that unexpected automation, security and privacy issues may be caused by such threats, which cannot be handled by existing IoT security mechanisms. To address this problem, we present HOMEGUARD, a system for appified IoT platforms to detect and cope with CAI threats. A symbolic executor module is built to precisely extract the automation semantics from IoT apps. The semantics of different IoT apps are then considered collectively to evaluate their interplay and discover CAI threats systematically. A user interface is presented to users during IoT app installation, interpreting the discovered threats to help them make decisions. We evaluate HOMEGUARD via a proof-of-concept implementation on Samsung SmartThings and discover many threat instances among apps in the SmartThings public repository. The evaluation shows that it is precise, effective and efficient.