The field of secure aggregation and confidential computing is moving towards the development of hybrid approaches that combine cryptography and trusted execution environments (TEEs) to improve performance and security. Researchers are exploring the use of TEEs, such as Arm Confidential Computing Architecture, to enable confidential and efficient machine learning applications, while also addressing the challenges of federated learning, including computation efficiency, attack tracing, and contribution assessment. Noteworthy papers include:

• A paper that introduces secure aggregation architectures integrating cryptographic and TEE-based techniques, analyzing trade-offs between security and performance.
• A paper that evaluates the performance-privacy trade-offs of deploying models within Arm Confidential Computing Architecture, showing promise for confidential and efficient ML applications.
• A paper that proposes a federated learning storage security model with homomorphic encryption to protect federated learning model privacy and address efficiency, attack tracing, and contribution assessment issues.