Data Poisoning CT-AI Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   What is the primary goal of a data poisoning attack on an AI-based system?

   • ATo corrupt the training data so that the model learns incorrect patterns or behaviors
   • BTo intercept data during transmission between the model and its users
   • CTo overload the model with excessive requests causing denial of service
   • DTo steal the model's parameters by querying it repeatedly
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   Which detection technique involves comparing the model's performance on a separate, trusted dataset to identify anomalies?

   • AValidation set checks
   • BFeature scaling
   • CModel compression
   • DData encryption
   Reveal Solution Discussion Select an answer first
4. 3application · medium

   A company is building a recommendation system. The training data is collected from user interactions on their website. A malicious actor creates many fake accounts and submits a large number of interactions to promote a specific product. Which attack vector and mitigation strategy are most relevant?

   • AAttack vector: data collection; Mitigation: implement user authentication and rate limiting to prevent fake account creation
   • BAttack vector: model training; Mitigation: use a more complex model to learn the patterns
   • CAttack vector: model deployment; Mitigation: monitor inference logs for anomalies
   • DAttack vector: preprocessing; Mitigation: apply feature scaling to normalize the data
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   Which of the following best describes the impact of data poisoning on a model's behavior?

   • AIt can cause the model to make systematic errors on specific inputs while performing normally on others
   • BIt has no impact on the model's behavior, only on the training time
   • CIt always causes the model to fail completely on all inputs
   • DIt only affects the model's inference speed, not its accuracy
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   During which phase of the AI development lifecycle can data poisoning occur?

   • AData collection and preprocessing
   • BOnly when the model is retrained in production
   • COnly during the model evaluation phase
   • DModel deployment and monitoring
   Reveal Solution Discussion Select an answer first