We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CR

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Cryptography and Security

Title: An Interpretable Federated Learning-based Network Intrusion Detection Framework

Abstract: Learning-based Network Intrusion Detection Systems (NIDSs) are widely deployed for defending various cyberattacks. Existing learning-based NIDS mainly uses Neural Network (NN) as a classifier that relies on the quality and quantity of cyberattack data. Such NN-based approaches are also hard to interpret for improving efficiency and scalability. In this paper, we design a new local-global computation paradigm, FEDFOREST, a novel learning-based NIDS by combining the interpretable Gradient Boosting Decision Tree (GBDT) and Federated Learning (FL) framework. Specifically, FEDFOREST is composed of multiple clients that extract local cyberattack data features for the server to train models and detect intrusions. A privacy-enhanced technology is also proposed in FEDFOREST to further defeat the privacy of the FL systems. Extensive experiments on 4 cyberattack datasets of different tasks demonstrate that FEDFOREST is effective, efficient, interpretable, and extendable. FEDFOREST ranks first in the collaborative learning and cybersecurity competition 2021 for Chinese college students.
Comments: 12 pages, draft
Subjects: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
Cite as: arXiv:2201.03134 [cs.CR]
  (or arXiv:2201.03134v1 [cs.CR] for this version)

Submission history

From: Tian Dong [view email]
[v1] Mon, 10 Jan 2022 02:12:32 GMT (1831kb,D)

Link back to: arXiv, form interface, contact.