BackAdversarial attack

Adversarial attack

Towards risk-aware artificial intelligence and machine learning systems: An overview

Zhang et al. (2022)

Sub-category

Risk Domain

2.2AI system security vulnerabilities and attacks

Vulnerabilities that can be exploited in AI systems, software development toolchains, and hardware, resulting in unauthorized access, data and privacy breaches, or system manipulation causing unsafe outputs or behavior.

"Recent advances have shown that a deep learning model with high predictive accuracy frequently misbehaves on adversarial examples [57,58]. In particular, a small perturbation to an input image, which is imperceptible to humans, could fool a well-trained deep learning model into making completely different predictions [23]."(p. 5)

Entity— Who or what caused the harm

Human

Due to a decision or action made by humans

AI system

Due to a decision or action made by an AI system

Other

Due to some other reason or is ambiguous

Intent— Whether the harm was intentional or accidental

Intentional

Due to an expected outcome from pursuing a goal

Unintentional

Due to an unexpected outcome from pursuing a goal

Other

Without clearly specifying the intentionality

Timing— Whether the risk is pre- or post-deployment

Pre-deployment

Occurring before the AI is deployed

Post-deployment

Occurring after the AI model has been trained and deployed

Other

Without a clearly specified time of occurrence

Supporting Evidence (1)

"In general, adversarial attacks can be grouped into two classes: 1. Targeted adversarial attack: The goal of targeted adversarial attack is to make an AI/ML model classify an adversarial image with a true label of K as a target class T (T ∕= K ) through intentional design (i.e., data manipulation). 2. Untargeted adversarial attack: The objective of untargeted adversarial attack is to make an AI/ML model generate a prediction that is different from the true label without intended target"(p. 5)