From the course: CompTIA SecAI+ (CY0-001) Cert Prep

Unlock this course with a free trial

Join today to access over 25,500 courses taught by industry experts.

Model skewing

Model skewing

From the course: CompTIA SecAI+ (CY0-001) Cert Prep

Start my 1-month free trial Buy for my team

Model skewing

“

Model skewing attacks occur when someone intentionally distorts an AI system's behavior over time by manipulating the data it consumes or the ways people use it. The attacker doesn't target the model's architecture or code. The attacker nudges the outputs in a specific direction. The change happens gradually and subtly and often avoids alerts. This attack works best in systems that update themselves with ongoing data such as feedback loops or reinforcement learning pipelines. the attacker doesn't need to poison the original training set. They feed the system enough biased, misleading, or carefully crafted malicious inputs so that the model's internal logic starts to drift. Consider a recommendation engine that learns from user clicks. An attacker repeatedly clicks on a specific type of content to push the system toward promoting a specific brand, ideology, or misinformation campaign. Over time, the model adapts to those signals and assumes that they reflect genuine interest. The…

Contents