The rapid proliferation of Internet of Things (IoT) devices has considerably improved user convenience while simultaneously raising significant privacy concerns. Existing privacy attacks typically focus on isolated threats, such as traffic classification, neglecting comprehensive inference of user behaviors. To overcome these limitations, we propose Isnooping, a novel end-to-end framework designed for traffic-based multidimensional privacy snooping attacks. Isnooping integrates traffic classification, user-personalized habitual behavior mining, and user action prediction. Consequently, it can accurately identify IoT devices and network applications, infer user-specific habitual behaviors, and anticipate future user actions, thereby exposing substantial privacy vulnerabilities. Extensive experiments demonstrate that Isnooping significantly outperforms current state-of-the-art methods.