AI Incident Response and Forensics COASP Practice Questions (Page 11)

Questions 51–55

1. 51foundation · easy

   What is the primary goal of attribution in AI incident response?

   • ATo delete all malicious data from the system
   • BTo identify the source and methods of the attack
   • CTo restore the AI system to normal operation
   • DTo contain the spread of the incident
   Reveal Solution Discussion Select an answer first
2. 52application · medium

   An e-commerce company's recommendation system has been manipulated to promote a specific vendor's products. The incident response team has confirmed that the model was poisoned during the training phase. The team needs to contain the incident while maintaining business continuity. Which containment strategy is MOST appropriate?

   • AApply a filter to the model's output to block recommendations from the specific vendor.
   • BRetrain the model immediately using the original clean dataset without analyzing the poisoning method.
   • CShut down the recommendation service entirely until a new model can be trained from scratch.
   • DDeploy the last known-good model version and route a percentage of traffic to it while the poisoned model is analyzed offline.
   Reveal Solution Discussion Select an answer first
4. 53application · medium

   A company's AI-powered chatbot has been manipulated to reveal sensitive customer information through a prompt injection attack. The incident response team is activated and must respond. Which action is MOST appropriate during the containment phase?

   • ANotify all customers whose information may have been exposed before completing the investigation.
   • BDisable the chatbot's access to the sensitive customer database while preserving the chatbot's logs and configuration for analysis.
   • CRetrain the chatbot with additional safety prompts to prevent future prompt injection attacks.
   • DImmediately delete the chatbot's conversation logs to prevent the sensitive information from being stored.
   Reveal Solution Discussion Select an answer first
5. 54application · medium

   During a forensic investigation of an AI incident, the team needs to preserve evidence from a model that was trained on sensitive customer data. Which evidence collection method is MOST appropriate?

   • ACreate a cryptographic hash of the model artifacts and training data, and store the hash in a secure location
   • BDelete the training data to protect customer privacy
   • CCopy the model artifacts to a shared network drive for easy access by the investigation team
   • DRely on the model's version control system to provide evidence
   Reveal Solution Discussion Select an answer first
6. 55foundation · easy

   Which of the following best describes the impact of a data poisoning attack on an AI model's integrity?

   • AThe model's inference speed decreases
   • BThe model becomes unavailable to users
   • CThe model's predictions become unreliable or maliciously biased
   • DThe model's source code is exposed to unauthorized parties
   Reveal Solution Discussion Select an answer first