期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2021
卷号：99
期号：14
语种：English
出版社：Journal of Theoretical and Applied
摘要：Computer networks constitute the vital artery of the information and communications technology era, allowing heterogeneous devices to communicate and share data. The immense number of Internet-connected devices with unpatched security vulnerabilities makes them susceptible to massive security attacks. Detecting unknown security attacks continues to be a major challenge, as they have been ranked constantly in the top three attack techniques since 2014. In this paper, the researchers aim to study the ability of supervised shallow and deep learning classifiers in detecting unknown botnet attacks on IoT devices. The performance of shallow and deep supervised learning classifiers was studied and compared using a well-known dataset (i.e., the Aposemat IoT-23 dataset). A thorough and extensive experimentation process was conducted (1000 experiments in total were performed), in which 12 unknown attack types and 38 unknown attack subtypes were studied under binary and multiclass classification problem. The results showed that the overall weighted average classification error rate was considerably high (61.46�86.40%), which dictates the importance of finding novel approaches and techniques to detect unknown attacks.
