A company has deployed an automated analytics platform across operations for tasks such as voice recognition. Which approach best addresses the risk that malicious users will degrade this platform’s outputs by altering the data used for training future models?
Turn off monitoring capabilities when the platform updates its models
Use one encryption process for every data set so training is uniform
Validate the training set and look for any suspicious inputs before use
Ensuring that training data is thoroughly validated and monitored for suspicious inputs is a strong defense against tampering. Logging controls or universal encryption processes do not directly address the risk of manipulated data that could undermine the platform’s predictive outputs, while deactivating monitoring further exposes the system to unauthorized modifications.
Ask Bash
Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.
What is data validation in the context of training machine learning models?
Open an interactive chat with Bash
How can malicious inputs degrade the performance of a machine learning model?
Open an interactive chat with Bash
Why are universal encryption processes insufficient for protecting training datasets from manipulation?