Offensive Artificial Intelligence: Current State of the Art and Future Directions

In recent years, the use of artificial intelligence (AI) has become increasingly prevalent in various industries, including cybersecurity. However, the same capabilities that make AI effective in protecting systems can also be used by attackers to launch sophisticated and complex cyberattacks. In this paper, the author conducted a systematic literature review to identify recent AI-driven cyberattacks and their characteristics, which could be used by cybersecurity experts to develop effective countermeasures. Preliminary findings indicate that some of the AI-driven attack types include data manipulation attacks (e.g. AI model data misclassification), access and penetration attacks (e.g. automated payload generation), and social engineering attacks (e.g. custom-made phishing). The paper also provides recommendations for policymakers, researchers, and practitioners to promote cybersecurity best practices, educate the public, and encourage international cooperation. Future research could delve deeper into the technical aspects of AI-driven cyberthreats and the strategies and techniques utilised to combat them. Overall, this paper emphasises the growing concern over offensive AI and adversarial AI and the need for continuous research and innovation to combat these evolving smart threats.