Encrypted Smartphone Traffic Reveals Behavioral Patterns Related to Sleep, Stress, and Loneliness
A team of researchers has shown that encrypted smartphone network traffic can be used to passively identify behavioral trends linked to sleep issues, stress, and feelings of loneliness. Utilizing a transformer backbone with user-specific adapters, the model captures both standard behaviors and anomalies. A sparse autoencoder is employed to derive understandable behavioral features from the encrypted information. Analysis through generalized estimating equations with Mundlak decomposition indicates that these three outcomes exhibit unique temporal patterns: stress is mainly tied to changes within individuals over time, whereas sleep disturbance and loneliness follow different trajectories. This study indicates that monitoring encrypted network traffic passively provides a scalable approach for ongoing behavioral tracking.
Key facts
- Encrypted smartphone network traffic can passively capture behavioral patterns related to sleep, stress, and loneliness.
- A transformer backbone with per-user adapters models shared behavioral structure.
- Sparse autoencoder extracts interpretable behavioral features.
- Generalized estimating equations with Mundlak decomposition separate between-person and within-person changes.
- Stress is primarily associated with within-person changes over time.
- Sleep disturbance and loneliness reflect different temporal structures.
- The study uses a ubiquitous, always-on, passive sensing modality.
- Research published on arXiv with ID 2605.01616.
Entities
Institutions
- arXiv