Artificial Intelligence in Cybersecurity
Abstract
As the world's infrastructure has quickly moved towards a Technological Revolution, so have the volume, velocity and sophistication of cyber threats. Though these are basics, traditional signature and rule based security solutions continue to be overwhelmed by polymorphism, zero day exploits and adapting adversarial tactics. The rapid evolution of technology has seen Artificial Intelligence (AI), including machine learning (ML), deep learning (DL), and natural language processing (NLP), become a game-changer in cybersecurity, creating the ability to learn from data, detect threats in real time, and act automatically to counter new attacks. This research paper reviews the literature on Artificial Intelligence in cybersecurity in a structured manner and analyzes literature obtained from the survey as well as literature obtained from secondary sources, and discusses the application of AI in cybersecurity, such as identification of intrusions, classification of malware, identification of phishing, prevention of fraud, and security orchestration. The research methodology used is a mixed research approach, which consists of a systematic literature review where the adoption trend and detection performance measure of the studies are quantitatively analyzed. It's found that the AI detection systems far surpass traditional detection systems in terms of both accuracy and reducing false-positives — but there are challenges associated with AI detection systems, including adversarial AI attacks, data privacy issues, explainability limitations and resource intensive model training. Overall, the paper concludes that, properly harnessed, AI can be a force multiplier to human expertise and significantly enhance organizational cyber resilience. Some directions for future research are also suggested — explainable AI (XAI), federated learning in the interest of a believable threat intelligence — and adversarial defense by AI versus AI itself.