A Survey on Adversarial Attacks and Defense Mechanisms in Artificial Intelligence

Adversarial attacks use the limitations of Artificial Intelligence models to generate mistakes and catastrophic failures of the system. They are a significant threat to Artificial Intelligence systems. Our objective in this work is to provide a systematic study of the different types of attacks we can encounter, their effects on AI systems, and the methods that are currently in place to combat them. We also discuss the nature of these attacks and how they impact the security of the AI model. Finally, we discuss the developments that need to be undertaken to make AI systems safer against these types of adversarial attacks.