DisenMamba: A New Framework for Network Traffic Anomaly Detection
A research paper on arXiv proposes DisenMamba, a novel disentangled multi-view Mamba framework for Network Traffic Anomaly Detection (NTAD). The authors identify a structural deficiency in existing multi-view Mamba scanning: redundancy accumulation, where view-invariant information is amplified and view-specific information diluted. DisenMamba reformulates multi-view scanning to address this issue. The paper is available at arXiv:2607.22829.
Key facts
- DisenMamba is a disentangled multi-view Mamba framework for NTAD.
- Existing multi-view Mamba scanning suffers from redundancy accumulation.
- Redundancy accumulation leads to representation homogenization and multi-view degradation.
- The paper is published on arXiv with ID 2607.22829.
- Network Traffic Anomaly Detection is a critical task in cybersecurity.
- Mamba offers linear-time complexity for long-sequence modeling.
- Multi-view scanning enhances detection precision through complementary contextual cues.
- DisenMamba reformulates multi-view scanning to solve redundancy accumulation.
Entities
Institutions
- arXiv