Robustness of AI systems - Poisoning Attacks and Mitigations

In this nugget, we will delve into the intriguing world of poisoning attacks within the context of machine learning security. First, you will gain a comprehensive understanding of what poisoning attacks entail, their mechanisms, and why they are a significant concern. Then, we will explore various types of poisoning attacks, aligning them with the existing taxonomy found in the literature. For each attack, we will provide a succinct explanation, shedding light on its unique characteristics and potential impact. Furthermore, we’ll discuss strategies and ideas for mitigating the adverse effects of these attacks, equipping you with valuable knowledge to bolster the security of machine learning systems against such threats.