ExclaveFL: Verifiable Claims for Federated Learning using Exclaves
Organizations: Imperial College London · Azure Research · Microsoft Security Response Center
Abstract
In federated learning (FL), data providers jointly train a machine learning model without sharing training data. This makes it challenging to provide verifiable claims about the trained model, e.g., related to the employed training data, data sanitization, or training algorithm---a malicious data provider can simply deviate from the correct training protocol without detection. Prior FL training systems have used trusted execution environments (TEEs) to protect the training computation, relying on the confidentiality and integrity of TEEs. TEE confidentiality, however, has been shown to be vulnerable to various attacks, e.g., via side channels. We describe ExclaveFL, a system for training FL models that establishes verifiable claims about models through fine-grained runtime attestation proofs. Since such proofs only require integrity, ExclaveFL generates them using exclaves, which are integrity-only execution environments without software-managed secrets, making them immune to data leakage attacks. ExclaveFL uses exclaves to attest data transformations during FL training, and the runtime attestation proofs then form an attested dataflow graph of the entire training computation. The graph is checked by an auditor to ensure that the trained model satisfies claims, e.g., about the use of data sanitization by data providers or aggregation by the model provider. Since exclaves require minor yet proprietary firmware changes, we evaluate a variant with similar performance but slightly weaker security guarantees on existing AMD SEV-SNP hardware. We show that ExclaveFL introduces less than 12% overhead compared to unprotected FL model training.
Figures & tables
| Exclave | Task API | Description |
|---|---|---|
| Training | StartTrain(...) | Receives global model |
| GetWeight(...) | Returns local model | |
| DP | AddNoise(...) | Performs local DP |
| Aggregation | Accept(...) | Receives local models |
| Aggregate(...) | Aggregates local models | |
| Model Update | ModelUpdate(...) | Applies diff to full model |
| GPT-2 | BERT | VGG9 | ||||
|---|---|---|---|---|---|---|
| Component | No Sec | ExFL | No Sec | ExFL | No Sec | ExFL |
| Training | 8,387 s | 8,659 s | 8,745 s | 9,164 s | 4,195 s | 4,641 s |
| DP | 1,035 s | 1,176 s | 699 s | 763 s | 20 s | 21 s |
| Aggregation | 9 s | 50 s | 6 s | 29 s | 1 s | 1 s |
| Model update | 5 s | 29 s | 2 s | 18 s | 1 s | 1 s |
| FL framework | 1,286 s | 1,983 s | 883 s | 1,257 s | 158 s | 124 s |
| #Clients | Signature | EDG | Total | Storage |
|---|---|---|---|---|
| #Rounds | Verification | Creation | ||
| 10 10 | 26 ms | 3 ms | 29 ms | <1 MB |
| 10 100 | 196 ms | 13 ms | 209 ms | 2 MB |
| 10 1,000 | 1,886 ms | 131 ms | 2,017 ms | 17 MB |
| 100 10 | 218 ms | 20 ms | 238 ms | 2 MB |
| 100 100 | 1,802 ms | 146 ms | 1,948 ms | 16 MB |
Federated Learning
Federated learning trains shared models across data holders without requiring them to pool their raw data.
See all 1,025 papers