Koga2022 Dataset: Comprehensive Dataset with Detailed Classification for Network Intrusion Detection Systems
Hideya Sato, Ryotaro Kobayashi · 2022
In this paper, we have created a communication dataset that addresses the problems of existing communication datasets and provides a fine-grained classification of malicious communications. Related research used publicly available datasets labeled normal and malicious to increase the accuracy of NIDS detection, as well as to study NIDS specific to certain attack scenarios. However, the existing dataset has an outdated creation date and only labels normal and malignant. Furthermore, even if attack scenarios that indicate broad policies for malicious communications are known, detailed classifications that indicate actual methods, etc., are few or unknown. To solve these problems, we created the Koga2022 Dataset, which is a detailed classification of malicious communication data collected using multiple penetration tools as well as a honeypot that is open to the outside world. This paper presents the statistics of the Koga2022 Dataset that was created. We then discuss validation methods that may be difficult with existing datasets and demonstrate the usefulness of the Koga2022 Dataset.