Akademik Veri Yönetim Sistemi
IEEE Transactions on Consumer Electronics, 2026 (SCI-Expanded, Scopus)
The rapid expansion of smart home device development has created considerable cybersecurity and privacy issues because of complicated interactions, heterogeneous data streams, and limited resources. To overcome these obstacles in the next-generation smart home setup, this paper proposed an edge-intelligent and privacy-conscious framework of anomaly detection based on adaptive threat learning and analysis (ATLAS). The model uses multi-modal representations by using tensor-graphs to learn cross-device interactions, predictive latent models to forecast temporal patterns, attention-based deviation scores with adaptive thresholds to identify abnormalities with high accuracy. The mechanisms of differential privacy and energy-aware computation provide security and efficiency of the framework to resource-constrained devices. Python-based extensive simulations prove that the proposed method was better in performance than state-of-the-art baselines, with high accuracy in recognition and detection, less latency and lower power demands, and better resistance to zero-day attacks without compromising privacy assurances. The findings demonstrate the performance of the framework in real-time, scalable, and privacy-aware anomaly detection in heterogeneous smart home systems.