How nativelogxd/repositorio-pipeline-kitsune works
Kitsune-py is an online network intrusion detector. Its main workflow accepts a packet capture or TSV, extracts packet features and incremental network statistics, then passes each feature vector through KitNET’s feature-mapping, training, and anomaly-scoring stages. The README also documents a Mirai demonstration that fits a log-score distribution and plots results. The sampled source supports the principal calls; unsampled feature-extraction details are represented only where the README or excerpts establish them.
Main entry points: FeatureExtractor.py Feature extractor · KitNET.py KitNET orchestrator · Kitsune.py Kitsune facade · netStat.py Traffic statistics
Components 12
Packet ingestion
- Feature extractor
FeatureExtractor.py
Network statistics
- Traffic statistics
netStat.py - Incremental statistics
AfterImage.py
Anomaly detection
- Kitsune facade
Kitsune.py - KitNET orchestrator
KitNET/KitNET.py - Feature clustering
KitNET/corClust.py - Denoising autoencoder
KitNET/dA.py - Model utilities
KitNET/utils.py - Demo and score analysis
example.py
Other components
- Capture or TSV
- tshark
- Scapy
Connections 11
- Capture or TSV → Feature extractor: provides packets
README.md - Feature extractor → tshark: parses capture
FeatureExtractor.py - Feature extractor → Scapy: parses fallback
FeatureExtractor.py - Feature extractor → Traffic statistics: updates statistics
FeatureExtractor.py - Traffic statistics → Incremental statistics: uses stream stats
netStat.py - Kitsune facade → Feature extractor: initializes and reads
Kitsune.py - Kitsune facade → KitNET orchestrator: processes vectors
Kitsune.py - KitNET orchestrator → Feature clustering: updates and clusters
KitNET/KitNET.py - KitNET orchestrator → Denoising autoencoder: trains and executes
KitNET/KitNET.py - Denoising autoencoder → Model utilities: uses activations
KitNET/dA.py - Demo and score analysis → Kitsune facade: runs packet loop
example.py