FreeUp: Frequency-Decoupled Anomaly Detection for Encrypted Traffic
A research paper from arXiv (2605.02970v1) introduces FreeUp, a frequency-decoupled framework for anomaly detection in encrypted network traffic. The authors identify a 'full-frequency' characteristic and 'spectral mismatch' limitation in existing image-based methods, which bias toward low-frequency information while encrypted traffic has prominent high-frequency components. FreeUp decomposes traffic data into distinct low- and high-frequency bands to address this mismatch and improve detection performance.
Key facts
- arXiv paper ID: 2605.02970v1
- Announce type: cross
- Focuses on encrypted network traffic anomaly detection
- Identifies 'full-frequency' characteristic and 'spectral mismatch' in current methods
- Proposes FreeUp framework
- FreeUp decomposes traffic into low- and high-frequency bands
- Aims to improve representation and detection performance
- Image-based methods are currently dominant for this task
Entities
Institutions
- arXiv