Figure 9.
Five heatmap panels depict S H A P value contributions of network traffic features to model output across instances, labelled A, B, C, D, and E.The visual depicts five panels labelled A, B, C, D, and E, each showing a heatmap of the S H A P value impact on model output across multiple instances on the X-axis. The Y-axis in each panel lists ranked network traffic features such as L 7 P r o t o, Protocol, Client T C P Flags, Shortest Flow P K T, Longest Flow P K T, Max I P P K T Len, Num P K T S Up To 128 Bytes, Num P K T S 128 To 256 Bytes, S r c To D s t A v g Throughput, S r c To D s t Second Bytes, In Bytes, T C P Win Max In, L 4 S r c Port, L 4 D s t Port, Server T C P Flags, and Sum of 32 other features. Each panel includes a horizontal trace labelled f x above the heatmap, indicating overall model output across instances. The colour scale represents negative to positive S H A P values, showing how individual feature contributions vary across instances, with each panel using a different contribution range while maintaining a consistent feature structure.

SHAP heatmap for NF-ToN-IoT-v2 data set for multiple class category (a–e) – benign, DDoS, DoS, reconnaissance, theft, respectively

or Create an Account

Close Modal
Close Modal