GraphIDS

Graph-based intrusion detection using GCN, Transformer autoencoder, and contrastive learning.

Reference implementation of the framework introduced in:

Govindarajan, V. & Muzamal, J. H. (2025). Advanced cloud intrusion detection framework using graph based features transformers and contrastive learning. Scientific Reports, 15, 20511. DOI: 10.1038/s41598-025-07956-w

pip install graphids

from graphids import GraphIDS

model = GraphIDS(n_features=41, n_classes=5)
model.train_pipeline(X_train, y_train)
result = model.evaluate(X_test, y_test)
print(f"Accuracy: {result.accuracy:.4f}")

Three-stage pipeline:

1. GCN — constructs a communication graph from flow data, extracts structural node embeddings via 3-layer graph convolution
2. Transformer autoencoder — refines embeddings through self-attention, identifies discriminative feature dimensions
3. Contrastive classifier — improves class separation for minority attack types (U2R, R2L), outputs multi-class predictions

@article{govindarajan2025graphids,
    title   = {Advanced cloud intrusion detection framework using graph based
               features transformers and contrastive learning},
    author  = {Govindarajan, Vijay and Muzamal, Junaid Hussain},
    journal = {Scientific Reports},
    volume  = {15},
    pages   = {20511},
    year    = {2025},
    doi     = {10.1038/s41598-025-07956-w},
}

MIT