Time series anomaly detection is crucial for improving the security and reliability of Cyber-Physical systems (CPS). While significant progress has been made, existing methods struggle to learn discriminative representations from multivariate time series with complex interactions and noise. To address this challenge, we propose a semi-supervised anomaly detection method based on denoising-aware contrastive learning, namely SSDCL, which can achieve robust performance for CPS anomaly detection using limited supervision. Specifically, we first design a similarity combination data augmentation algorithm to handle complex interactions among continuous sensor measurements and discrete actuator states. Furthermore, we develop a denoising hierarchical contrastive loss function that mitigates data noise interference while ensuring discriminative spatio-temporal representation. To validate the effectiveness of SSDCL, we conducted empirical evaluations on three publicly available CPS time series datasets including PUMP, SWaT and WADI. The experimental results show that the proposed method achieves F1 Score of 97.5%, 93.0%, and 74.4%, respectively, outperforming the state-of-the-art (SOTA) CPS anomaly detection methods.

Alzheimer’s disease (AD) is the leading cause of dementia in the elderly, and its numbers are rising rapidly. In recent years, many researchers have used convolution neural networks and vision transformer to classify AD. However, most networks do not have a good balance between classification performance and efficiency. To solve this problem, this paper proposes a new Light-Patch and Dynamic window local attention Vision Transformer network (LP-DWLA-ViT) to classify AD. The network includes a Light-Patch (LP) module and a Dynamic window local attention (DWLA) module. LP module uses convolution layer and smaller patch, reduces computation and improves classification efficiency. DWLA module achieves both classification performance and efficiency by dividing the attention computing window and dynamically changing the window size. The LP-DWLA-ViT network has been extensively tested on ADNI datasets. Its accuracy up to 99.36%, specificity up to 99.71% and sensitivity up to 99.46%.