ISACA AI Fundamentals Practice Test

Differential privacy is the technique that adds random noise to data or to the results of data queries to protect individual privacy. By injecting carefully calibrated randomness, it ensures that the presence or absence of a single person’s data has only a tiny effect on the overall output. This provides formal privacy guarantees, often described with a privacy parameter that controls the trade-off between privacy strength and data utility: smaller values mean stronger privacy but noisier results.

In practice, methods like the Laplace or Gaussian mechanisms add noise proportional to a query’s sensitivity, so aggregate statistics (like averages or counts) remain useful while making it hard to infer information about any one individual. The noise can be designed to hold up under repeated analyses, limiting how much private information can be leaked over time.

Other terms are not techniques for protecting privacy through data release. Deepfakes involve creating realistic synthetic media, while GDPR and CCPA are regulatory frameworks that set rules for how data can be collected, stored, and processed rather than methods for protecting privacy within data outputs.