Secure aggregation enhances the security of federated learning systems by:

Question

Options

• AProcessing client updates in isolation to reduce the risk of exposing sensitive information
• BApplying differential privacy techniques to mask sensitive information in training data
• CEncrypting individual model updates during transmission to ensure only the server can access the
• DEnsuring individual client contributions remain confidential even if the server is compromised

• A
  7% (2)
• B
  4% (1)
• D
  89% (24)

Explanation

Secure aggregation uses cryptographic protocols (such as secret sharing or homomorphic encryption) so that the central aggregation server can only compute the sum or average of all client model updates - it cannot see any individual client's update. This protects client data even if the server itself is compromised or malicious. Option A describes isolation of updates, which is a different concept. Option B describes differential privacy, a separate privacy-enhancing technique. Option C describes encryption in transit, which only protects data during transmission; secure aggregation goes further by protecting individual updates from the server itself, even after receipt.

No community discussion yet for this question.