Few Edges Are Enough: Few-Shot Network Attack Detection with Graph Neural Networks
Tristan Bilot, Nour El Madhoun, Khaldoun Al Agha, Anis Zouaoui · 2025
Read Few Edges Are Enough: Few-Shot Network Attack Detection with Graph Neural Networks by Tristan Bilot, Nour El Madhoun, Khaldoun Al Agha, Anis Zouaoui (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-16964.txt.
Description
Bilot and three co-authors built a network attack detector that learns from almost no labeled attacks in twenty twenty-five. Graph neural networks model network traffic as a graph and detect attacks well, but labeled examples are scarce in practice. Unsupervised methods flag anomalies yet confuse them with true attacks.
Their system, FEAE, trains with self-supervised learning plus few-shot learning. A hybrid objective mixes contrastive and reconstruction signals, teaching the network normal traffic structure without labels. Then a handful of labeled attack events, drawn as edges in the graph, teach it to separate real attacks from odd but harmless traffic. On two standard datasets it matched supervised and unsupervised rivals. Just one example per attack type brought substantial gains, and on one dataset it beat some fully supervised systems.
Catalog record
| CREATOR | Tristan Bilot, Nour El Madhoun, Khaldoun Al Agha, Anis Zouaoui |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-16964