A Realistic Experimental Comparison of the Suricata and Snort Intrusion-Detection Systems

Abstract

The Suricata intrusion-detection system for computer-network monitoring has been advanced as an open-source improvement on the popular Snort system that has been available for over a decade. Suricata includes multi-threading to improve processing speed beyond Snort. Previous work comparing the two products has not used a real-world setting. We did this and evaluated the speed, memory requirements, and accuracy of the detection engines in three kinds of experiments: (1) on the full traffic of our school as observed on its "backbone" in real time; (2) on a supercomputer with packets recorded from the backbone; and (3) in response to malicious packets sent by a red-teaming product. We used the same set of rules for both products with a few small exceptions where capabilities were missing. We conclude that Suricata can handle larger volumes of traffic than Snort with similar accuracy, and that its performance scaled roughly linearly with the number of processors up to 48. We observed no significant speed or accuracy advantage of Suricata over Snort in its current state, but it is still being developed. Our methodology should be useful for comparing other intrusiondetection products.