Cross-Dataset Generalization of Few-Shot NIDS

Determine the robustness of few-shot network intrusion detection approaches to unseen network environments by training and evaluating representative methods on disjoint datasets.

Background

The reviewed studies evaluate their approaches within individual datasets, but none trains on one dataset and evaluates on a disjoint dataset. Consequently, the literature does not establish whether few-shot network intrusion detection methods generalize beyond the network environments represented in their training data. Cross-dataset evaluation is therefore needed to assess robustness to unseen environments and to complement the proposed unified benchmarking and reproducibility efforts.

References

Second, generalization across datasets remains essentially untested: none of our included studies trains and evaluates on disjoint datasets, so the robustness of these approaches to unseen network environments is unknown.

— Few-Shot Learning for Network Intrusion Detection: Methods, Datasets, and Performance  (2609.11275 - Roszeitis et al., 10 Sep 2026) in Section 5, subsection “Future Work”