Google Professional Cloud Security Engineer Question 368
Select 4Google Cloud PlatformYou are a Cloud Security Engineer tasked with securing an AI/ML system on Google Cloud that processes sensitive customer data. To prevent unintentional exploitation of the data or model, which of the following actions should you take?
- A
Implement differential privacy techniques to ensure results do not reveal individual data points.
- B
Enable Vertex AI Model Monitoring to detect data drift and anomalies in model predictions.
- C
Encrypt sensitive data used for training using Cloud KMS before uploading it to Cloud Storage.
- D
Disable IAM permissions for all users to prevent access to training data.
- E
Use Explainable AI to analyze and mitigate potential biases in the AI/ML model.
Show answer and explanation
Correct answers: A, B, C, E
Explanation
Preventing unintentional exploitation of AI/ML systems requires a combination of data security measures, model monitoring, and bias analysis. Differential privacy protects individual data points, encryption ensures data security, and Vertex AI Model Monitoring and Explainable AI help safeguard the integrity and fairness of the model. Disabling all IAM permissions is not a viable approach, as it prevents legitimate access and usage of the system.
- A. Correct.
Differential privacy ensures that AI/ML outputs do not reveal sensitive information about individual data points, which is crucial for preventing unintentional exploitation of customer data.
- B. Correct.
Vertex AI Model Monitoring helps detect and address data drift and anomalies, which could otherwise lead to inaccurate predictions and potential security vulnerabilities.
- C. Correct.
Encrypting sensitive data using Cloud KMS before uploading to Cloud Storage ensures that the data remains secure during storage and transit, mitigating the risk of unauthorized access.
- D. Incorrect.
Disabling IAM permissions for all users is overly restrictive and impractical. Instead, permissions should be managed using the principle of least privilege to ensure secure access.
- E. Correct.
Explainable AI can help identify and mitigate biases or vulnerabilities in the model, reducing the risk of exploitation or unintended consequences.