Work

人と協調してネットワーク運用を支援する機械学習に関する研究

Year: 2023
Type: dissertation
Abstract: Communication networks have become indispensable to people's lives.In this age of smartphones, the Internet can be regarded as an infrastructure for daily life and electricity, water, and gas.Companies that provide such communication services to users struggle daily to ensure the stable communication networks.Network operations experts are needed to ensure the stable communication networks.However, with new technologies such as the Internet of things (IoT) and 5G, the burden on the experts who manage them continues to increase.The work of network operations experts is diverse and includes many decision-making tasks.We focus on the intrusion detection and prevention system (IDPS) signature classification task in network operations.IDPSs monitor network systems and take actions such as logging, notification, and blocking when malicious communications are detected.This dissertation focuses on a type of IDPS that performs detection based on pattern files of malicious communications such as signatures.The signatures are distributed periodically by the IDPS developers, similar to a subscription service.Network operations experts determine the importance level ("low", "medium", or "high") of each signature to set IDPS actions.For example, if the importance of a signature is "high", the action is "blocking"; if it is low, the action is "logging".Determining the importance level of a signature for setting it up in the IDPS is a burden for the expert.While expert-designed if-then rule scripts can automatically determine some signatures, the remaining signatures must be determined manually by experts based on elements in the signatures, articles on the Internet, and their own experience.This manual decision-making process takes some time.In addition to time consuming issues, it takes sufficient knowledge and experience to determine the importance level, and there are not many experts who have such knowledge and experience.In other words, the cost of hiring experts is high.Therefore, determining the importance of a signature, which requires an expert's time, is also a significant cost to the network company.IDPS signature classification is an important task, but there has yet to be research to automate it.We must recognize the seriousness of classification errors that can result from automation, as is the case in the medical field.Signature misclassification leads to IDPS misconfiguration.Misconfiguration of IDPS can cause security incidents, such as missing malicious communications and false interceptions of regular communications.Security incidents should be avoided because they damage public trust in the organization operating the network system.Hence, it is not practical to automate all signature classifications.In order to automate
Author: 川口, 英俊
Language: English
Open Access status: green
OpenAlex ID: W7016726042
Cites: 104
Cited by:
Related to: 9