Artificial Intelligence Based Cybersecurity: Two-Step Suitability Test
Shah Md. Istiaque, Md Toki Tahmid, Asif Iqbal Khan, Zaber Al Hassan, Sajjad Waheed · 2021
Enormous network connectivity, big data, the Internet of Things (IoT), digitalization of the world, and the use of social websites and apps have brought enormous institutional and individual security challenges. The conventional security system often fails to provide cyber security to institutions and individuals. Artificial Intelligence (AI) is highly adaptive and smart to handle the volatile cyber security environment. AI plays a prudent role in access control, user authentication and behavior analysis, spam, malware, and botnet detection. Machine learning (ML) models are the building blocks of AI. In this research, a novel practical approach is followed to prove the effectiveness of AI in the field of cyber-security. Multiple machine learning algorithms are applied to prove that. A two-step suitability test is conducted in this study. In the 1st step, the KDD'99 data set is used to train and test the AI models. In the second step, train models are repeatedly tested on a fresh data set, NSL-KDD. Finally, the testing results are compared to prove the authenticity of AI in cyber-security. An absolute practical approach and reliable outputs of various AI models prove AI's suitability in cyber-security.