.png){kind=icon}
.png){kind=icon}
Return
MP-DEC: multi-perspective deep embedding custering for network protocol data
.png){kind=icon}
PRE
AI
DOI：10.1080/17445760.2026.2620536.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
To address rapidly evolving cyber threats, precise protocol analysis is fundamental to modern defenses. This paper proposes MP-DEC: a multi-perspective deep clustering framework for hierarchical protocol analysis. Its innovations include a feature weighting mechanism that distinguishes protocol types from formats, a featureenhanced autoencoder integrating these weights, and a joint optimization strategy enabling precise hierarchical separation. Evaluated on curated and MACCDC 2012 datasets, MP-DEC achieves 91.8% V-Measure for protocol types and 93.6% clustering purity for formats. The framework provides more interpretable, structured representations for the tested structured non-encrypted protocol data, offering foundational support for downstream security tasks like anomaly detection and threat hunting.
Keywords:
Cybersecurity
protocol clustering
deep embedding clustering
autoencoder
network traffic analysis
Journal
I
IF:
0.7
Papers:
45
Citations:
266
Organization
No organization information available