Attacking the Source
Underlying Technology Behind AI System
The underlying technology behind AI systems is software and algorithms.
Vulnerabilities in AI Systems
AI systems vulnerabilities are often input handling and logic flaws.
Unexpected and undesired AI behaviors
Not providing a large amount of trained data can lead to unexpected and undesired AI behaviors.
Data poisoning
Data poisoning is when trained data has been tainted.
Evasion attack
Evasion attacks are designed to interrupt AI’s ability to perform classification and identification.
Inference attack
Inference attacks can be introduced when a specific connection in an AI system can be unraveled and detected.
Extraction attack
An extraction attack is when a malicious actor feeds data into a model and tracks how the model manipulates the data.
