Authentication technology is a key area of research in Human-Computer Interaction. Although facial and gait recognition technologies have achieved significant results, these methods may pose potential threats to user privacy, and the high cost of equipment limits their popularity in certain scenarios. In contrast, WiFi sensing has received widespread attention in academia and industry due to its non-intrusiveness, privacy protection, and low-cost deployment. WiFi-based person identification is typically achieved by analyzing Channel State Information in WiFi signals. Most WiFi-based identification methods rely on gait, which is expensive to train, requires strong cooperation from users, and faces performance challenges, especially when the user scale increases. To overcome these limitations, we propose a WiFi-based authentication system, WiLDID, which enables seamless identification without requiring excessive user cooperation. WiLDID introduces an end-to-end lightweight deep neural network, directly extracting features from raw WiFi signals, avoiding complex signal preprocessing. The backbone network improves upon StarNet, effectively extracting high-dimensional implicit features while reducing model size and parameter count. Experimental results show that the system achieves a accuracy of 98% in tests involving 36 individuals. We also implemented real-time authentication, further validating the feasibility and efficiency of WiFi-based static identity recognition in Internet of Things.