Learn about poisoning attacks in the context of AI. Prepare for the IBM AI Fundamentals certification exam with this comprehensive explanation.
Table of Contents
Question
Which of the following is an example of poisoning in the AI context?
Select the three that apply.
A. Updating features and labels of the training data
B. Corrupting test data sent to the deployed model
C. Sending malicious test samples to the deployed model
D. Injecting malicious samples into the training data
E. Modifying AI model architecture, parameters, and logic
Answer
A. Updating features and labels of the training data
D. Injecting malicious samples into the training data
E. Modifying AI model architecture, parameters, and logic
Explanation
Poisoning refers to corrupting the training set in a way that causes an AI model to produce inaccurate results.
In the context of AI, poisoning refers to the act of introducing harmful data or alterations that can negatively affect the model’s learning process or performance. The examples of poisoning in AI include:
A. Updating features and labels of the training data: This can mislead the AI during its learning phase, causing it to learn incorrect patterns.
D. Injecting malicious samples into the training data: Similar to updating features and labels, injecting malicious samples can corrupt the training process.
E. Modifying AI model architecture, parameters, and logic: Altering the fundamental aspects of an AI model can lead to it learning or functioning incorrectly.
Options B and C, which involve corrupting test data or sending malicious samples to the deployed model, are more indicative of an adversarial attack on an already trained and deployed model, rather than poisoning which occurs during the training phase.
IBM Artificial Intelligence Fundamentals certification exam practice question and answer (Q&A) dump with detail explanation and reference available free, helpful to pass the Artificial Intelligence Fundamentals graded quizzes and final assessments, earn IBM Artificial Intelligence Fundamentals digital credential and badge.