During red-team testing of an AI system used to make lending decisions, which of the following techniques BEST simulates a data poisoning attack?

The correct answer is D. Corrupting training data sets to manipulate outcomes. Data poisoning is an attack where an adversary corrupts or manipulates the training data used to build an AI model, causing the model to learn incorrect patterns and produce biased or malicious outcomes. Option D directly describes this: corrupting training datasets to…

AI Security Assurance and Resilience

Question

Options

AInputting encrypted data into the model
BAdding noise to output predictions
CStealing model weights from a deployed API
DCorrupting training data sets to manipulate outcomes

How the community answered

(25 responses)

A
8% (2)
B
4% (1)
D
88% (22)

Explanation

Data poisoning is an attack where an adversary corrupts or manipulates the training data used to build an AI model, causing the model to learn incorrect patterns and produce biased or malicious outcomes. Option D directly describes this: corrupting training datasets to manipulate outcomes - for example, injecting fraudulent loan records labeled as 'low risk' to make the model approve bad applicants. Option A (encrypting input data) affects inference, not training. Option B (adding noise to predictions) is output manipulation, not a training-phase attack. Option C (stealing model weights) is a model theft/exfiltration attack, not poisoning.

Topics

#Data poisoning#Red team testing#AI attack types#AI security testing

Community Discussion

No community discussion yet for this question.

Full AAISM Practice