Analyze the User Behaviors via Network Traffic Data in Smart Home

As a service medium for household units, smart homes can serve as energy management hubs by analyzing user behavior within the home to optimize energy allocation and usage, thereby avoiding unnecessary resource waste—for example, turning off the TV while the user is taking a shower. However, many users refuse to allow third-party analysis of their smart device data due to privacy concerns. To address this issue, we propose a technique for accurately extracting user behavior features from network traffic, enabling non-intrusive user behavior recognition. Recognizing user behavior poses significant challenges due to the uncertainty of user activities and the encrypted nature of network traffic. To overcome these challenges, we leverage a dynamic time warping (DTW) algorithm to transform encrypted traffic data into device event sequences using device event fingerprints, facilitating the matching of user activity patterns. Furthermore, we introduce a deterministic finite automaton (DFA) model to capture transition rules within activity sequences, enhancing recognition accuracy. We evaluated our approach on a self-constructed real-world dataset, and the experimental results demonstrate that our method achieves a behavior recognition accuracy of up to 87.3%.