How zhiyu-l/csc3104-honeypot-fl works
This repository implements federated honeypot intrusion detection: Cowrie sessions become labelled features, organisations train locally, and a coordinator aggregates model updates into versioned global models. A parallel dataset workflow splits training and replay data, then evaluates models. The README also documents nightly feature ingestion and a model API, though their implementation is not represented in the sampled source. Edges are limited to documented workflows and relationships visible in excerpts or the source index.
Main entry points: server_app.py Coordinator app · process_replay.py Replay processing · task.py Local training · cowrie_sessions.py Session labeling
Components 22
Data preparation
- Cowrie honeypots
services/cowrie - Session labeling
tools/cowrie_dataset/cowrie_sessions.py - Org data splits
tools/cowrie_dataset/split_orgs.py - Replay test data
tools/cowrie_dataset/replay_testset.py
Feature processing
- Replay processing
services/log-processor/process_replay.py - Nightly processing
services/log-processor/nightly.py - Feature batches
services/client/src/fl_client/data.py
Federated learning
- Client app
services/client/src/fl_client/client_app.py - Local training
services/client/src/fl_client/task.py - Coordinator app
services/coordinator/src/fl_coordinator/server_app.py - Fault-tolerant rounds
services/coordinator/src/fl_coordinator/runner.py - Aggregation strategies
services/coordinator/src/fl_coordinator/strategies.py - Adaptive aggregation
services/coordinator/src/fl_coordinator/adaptive.py - Run checkpoints
services/coordinator/src/fl_coordinator/checkpoints.py - Model promotion
services/coordinator/src/fl_coordinator/publish.py
Shared model layer
- Feature schema
libs/fl-common/src/fl_common/schema.py - Shared classifier
libs/fl-common/src/fl_common/model.py - Model evaluation
libs/fl-common/src/fl_common/evaluation.py - Synthetic data
libs/fl-common/src/fl_common/synthetic.py
Model serving
- Model API
services/model-api
Other components
- Pipeline operator
- SSH attacker
Connections 29
- SSH attacker → Cowrie honeypots: sends traffic
README.md - Cowrie honeypots → Session labeling: produces logs
README.md - Session labeling → Org data splits: labels sessions
README.md - Session labeling → Replay processing: shares extractor
services/log-processor/process_replay.py - Org data splits → Local training: provides training data
README.md - Org data splits → Replay test data: provides held-out data
README.md - Replay test data → Replay processing: selects replay sessions
tools/cowrie_dataset/replay_testset.py - Replay processing → Feature schema: uses schema
services/log-processor/process_replay.py - Replay processing → Feature batches: writes feature batches
README.md - Nightly processing → Replay processing: invokes processing
services/log-processor/nightly.py - Nightly processing → Feature batches: publishes batches
README.md - Feature batches → Local training: loads data
services/client/src/fl_client/task.py - Client app → Local training: trains locally
services/client/src/fl_client/client_app.py - Client app → Shared classifier: loads weights
services/client/src/fl_client/client_app.py - Local training → Feature schema: validates features
services/client/src/fl_client/task.py - Local training → Synthetic data: uses fallback data
services/client/src/fl_client/task.py - Coordinator app → Shared classifier: initializes model
services/coordinator/src/fl_coordinator/server_app.py - Coordinator app → Aggregation strategies: selects strategy
services/coordinator/src/fl_coordinator/server_app.py - Coordinator app → Fault-tolerant rounds: runs rounds
services/coordinator/src/fl_coordinator/server_app.py - Fault-tolerant rounds → Aggregation strategies: aggregates updates
services/coordinator/src/fl_coordinator/runner.py - Aggregation strategies → Adaptive aggregation: uses adaptive logic
services/coordinator/src/fl_coordinator/strategies.py - Fault-tolerant rounds → Client app: dispatches rounds
- Coordinator app → Run checkpoints: saves run
services/coordinator/src/fl_coordinator/server_app.py - Coordinator app → Model promotion: promotes model
services/coordinator/src/fl_coordinator/server_app.py - Coordinator app → Model evaluation: evaluates final model
services/coordinator/src/fl_coordinator/server_app.py - Pipeline operator → Session labeling: runs pipeline
README.md - Pipeline operator → Coordinator app: starts training
README.md - Pipeline operator → Model evaluation: tests model
README.md - Model promotion → Model API: publishes versions