Few-Shot Learning for Network Intrusion Detection: Methods, Datasets, and Performance
A PRISMA-style review of 21 few-shot learning studies for network intrusion detection finds meta-learning and CNNs dominant and evaluation inconsistently reported.
The systematic review screened 1,358 records from ACM Digital Library, IEEE Xplore, and Scopus covering 2022-2026 and retained 21 studies on few-shot learning for network intrusion detection. Meta-learning (8 studies) and convolutional neural networks (10) are the most common approaches, while CIC-IDS2017 and CSE-CIC-IDS2018 are the most frequently used datasets. Most evaluations use five or fewer samples per class, and missing parameters and source code limit reproducibility and direct comparison.
- 21 FSL-for-NIDS studies retained from 1,358 initial records
- CNNs (10) and meta-learning (8) dominate the approaches
- CIC-IDS2017 and CSE-CIC-IDS2018 are the most used benchmarks
- Most settings evaluate five or fewer samples per class
- Reproducibility hindered by missing parameters and source code
- Products
- CIC-IDS2017CSE-CIC-IDS2018
From the original article
1 min read at arxiv.orgAnomaly-based network intrusion detection systems (NIDS) are an important first line of defense. However, training NIDS for new attack types is challenging, because labeled attack data are rarely available. Few-shot learning (FSL) addresses this problem by learning from few samples.
By Arne Roszeitis · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.