Abstract: The disclosed embodiments relate to a system that anonymizes sensor data to facilitate machine-learning training operations without disclosing an associated user's identity. During operation, the system receives encrypted sensor data at a gateway server, wherein the encrypted sensor data includes a client identifier corresponding to an associated user or client device. Next, the system moves the encrypted sensor data into a secure enclave. The secure enclave then: decrypts the encrypted sensor data; replaces the client identifier with an anonymized identifier to produce anonymized sensor data; and communicates the anonymized sensor data to a machine-learning system. Finally, the machine-learning system: uses the anonymized sensor data to train a model to perform a recognition operation, and uses the trained model to perform the recognition operation on subsequently received sensor data.

Abstract: The disclosed embodiments relate to a system that anonymizes sensor data to facilitate machine-learning training operations without disclosing an associated user's identity. During operation, the system receives encrypted sensor data at a gateway server, wherein the encrypted sensor data includes a client identifier corresponding to an associated user or client device. Next, the system moves the encrypted sensor data into a secure enclave. The secure enclave then: decrypts the encrypted sensor data; replaces the client identifier with an anonymized identifier to produce anonymized sensor data; and communicates the anonymized sensor data to a machine-learning system. Finally, the machine-learning system: uses the anonymized sensor data to train a model to perform a recognition operation, and uses the trained model to perform the recognition operation on subsequently received sensor data.