Secure Aggregation for One-Shot Clients Without Synchronization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing secure aggregation protocols face challenges such as synchronization latency, integrity verification complexity, scalability issues, and coordination difficulties in distributed systems, particularly when dealing with large numbers of clients or clients with intermittent connectivity.
Innovation Solution
A one-shot client participation model using key-additive homomorphic encryption (KAHE) and additive homomorphic encryption (AHE) schemes, combined with a distributed decryption mechanism, allows clients to submit inputs without synchronization and ensures secure aggregation with minimal computational overhead and verification mechanisms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional secure aggregation protocols are used with multiple synchronization rounds, then security and integrity verification can be achieved, but protocol latency increases and scalability deteriorates
Solution Approach 1:
The server performs preliminary actions by pre-computing and storing aggregation proofs for previous client inputs before receiving new inputs. This allows the server to verify integrity and prepare aggregation results in advance, eliminating the need for synchronized verification rounds and reducing protocol latency while maintaining security.
Solution Approach 2:
The protocol segments the aggregation process into independent client submissions and server processing phases. Each client submits encrypted inputs asynchronously without waiting for others, and the server handles aggregation and verification separately, eliminating synchronization dependencies and reducing overall protocol time.
2Ease of operation
If traditional secure aggregation protocols require client synchronization, then coordination and verification can be managed, but scalability to large numbers of clients deteriorates
Solution Approach 1:
Each client independently computes and submits their own encrypted input along with an aggregation proof without requiring coordination with other clients. The server independently verifies each submission using pre-stored proofs and aggregates results autonomously, eliminating the need for client synchronization and enabling linear scalability with the number of clients.
Solution Approach 2:
The server pre-computes and stores aggregation proofs for previous client inputs before receiving new submissions. This preliminary preparation allows the server to verify integrity and aggregate results independently for each client submission, eliminating coordination overhead and enabling efficient scaling to large numbers of clients.
3Reliability
If distributed decryption is used with multiple decryptors, then reliability and fault tolerance improve, but system complexity and computational overhead increase
Solution Approach 1:
The patent extracts the decryption function from the main aggregation process by using a separate key management system. The server stores encrypted client inputs and their corresponding decryption keys securely, allowing authorized parties to decrypt individual inputs or the final aggregate without complex distributed decryption protocols, thus reducing system complexity while maintaining reliability.
Solution Approach 2:
The server acts as an intermediary that manages the decryption process by storing encrypted inputs and keys securely. Instead of requiring multiple decryptors to coordinate, the server mediates by providing decrypted inputs or the final aggregate to authorized parties, simplifying the system architecture while maintaining fault tolerance through secure key management.
Data Source
AI summary
Methods and systems for implementing secure aggregation with one-shot clients are described herein. A server receives, from each client, (i) an encrypted client input represented by a client input encrypted by a Key-Additive Homomorphic Encryption (KAHE) scheme using a client key, and (ii) an encrypted client key represented by the client key encrypted by an Additive Homomorphic Encryption (AHE) scheme using a public key received by the client from a decryptor. The server adds the encrypted client input to a combination (e.g., a running sum) of encrypted client inputs received from at least some of the clients. The server further adds the encrypted client key to a combination (e.g., a running sum) of encrypted client keys received from the clients which supplied their client inputs to the server. The server then transmits, to the decryptor, the running sum of encrypted client keys. In response, the server receives, from the decryptor, a decrypted key produced by decrypting, using a secret key corresponding to the public key, the running sum of encrypted client keys. The server then decrypts, using the decrypted key, the running sum of encrypted client inputs.