Apple is addressing the challenge of combining impressive yet data-intensive AI features with privacy-friendly practices. The company plans to tackle this issue using a unique type of encryption known as homomorphic encryption. In a statement from October 24, Apple explained its approach to integrating machine learning with homomorphic encryption to process data confidentially. Some features based on this technology are already available in iOS 18.
Homomorphic encryption (HE) allows computations to be performed on data while it remains encrypted. This means private data is not processed locally on the device. Instead, it is sent encrypted to Apple servers, where it is processed. HE maintains certain structures of the data, ensuring that computations on encrypted data yield a correct but still encrypted result. Only the original device can decrypt this result, preventing Apple or third parties from accessing the data or the outcome.
The concept of homomorphic encryption is not new, but it has historically suffered from significant inefficiencies. Theoretically functional systems were so inefficient that practical use of homomorphically encrypted computations was either impossible or not economically viable. However, recent advancements in new HE encryption systems and improved hardware acceleration have reduced these inefficiencies, making HE increasingly practical.
For example, Microsoft’s Edge browser has, for several years, used a confidential security check based on APSI (Asymmetric Private Set Intersection). This technology compares locally stored website login data against public leak databases without needing to store these large databases locally or reveal login data to the comparison server. Homomorphic encryption, specifically Microsoft’s open-source SEAL implementation, provides the technical foundation for this feature.
While Apple’s initiative with homomorphic encryption is notable, it is part of a broader trend in the tech industry to enhance data privacy without compromising functionality. As more companies adopt similar privacy-focused technologies, users can expect to see a growing number of features that protect their data while delivering advanced capabilities.
Overall, the integration of homomorphic encryption into Apple’s services represents a significant step towards reconciling the demands of cutting-edge AI features with the need for robust data privacy. As technology continues to evolve, it is likely that we will see further innovations in this area, helping to ensure that personal data remains secure in an increasingly digital world.