Intel’s Heracles Chip Accelerates Fully Homomorphic Encryption

Concerns about data privacy are driving innovation in secure computation. Fully homomorphic encryption (FHE), a method allowing computation on encrypted data without decryption, promises a solution. While, the intensive processing demands of FHE have historically limited its practicality. Now, Intel is demonstrating significant progress with its Heracles chip, achieving up to a 5,000-fold speedup in FHE tasks compared to standard server CPUs .

The Challenge of Fully Homomorphic Encryption

FHE enables secure computation by allowing operations to be performed directly on encrypted data. This eliminates the need to decrypt sensitive information before processing, addressing privacy concerns in applications like cloud computing, healthcare, and financial services . However, the computational overhead associated with FHE has been substantial. Traditional CPUs and GPUs struggle to efficiently handle the complex mathematical operations required, making FHE significantly slower than processing unencrypted data.

Intel’s Heracles: A Dedicated FHE Accelerator

To overcome this performance bottleneck, Intel developed Heracles, a specialized hardware accelerator designed specifically for FHE workloads. Demonstrated at the IEEE International Solid-State Circuits Conference (ISSCC) in San Francisco, Heracles utilizes a 3-nanometer FinFET technology and boasts a size approximately 20 times larger than other FHE research chips. It’s also equipped with two 24-gigabyte high-bandwidth memory chips, a configuration typically found in GPUs used for AI training .

Real-World Performance Gains

Intel showcased Heracles’ capabilities with a practical example: a private ballot verification system. In this scenario, a voter encrypts their ID and vote and sends it to an encrypted database. The system verifies the ballot without decrypting the information, returning an encrypted response that the voter can then decrypt.

The results were dramatic. A process that took 15 milliseconds on an Intel Xeon server CPU was completed in just 14 microseconds by Heracles. Extrapolating this performance gain, verifying 100 million voter ballots would grab over 17 days on a standard CPU, but only 23 minutes with Heracles .

FHE vs. Secure Multi-Party Computation (MPC)

FHE is often compared to secure multi-party computation (MPC). While both aim to protect data during computation, they differ in their approach. MPC involves distributing data among multiple parties and performing computations interactively. FHE, protects data through encryption and allows computation without interaction, with the decryption key potentially being shared securely .

The Future of FHE

While Intel believes Heracles represents a significant leap forward, other startups are also actively developing FHE accelerators. Sanu Mathew, Intel’s security circuits research lead, asserts that Heracles’ scale and compute performance give Intel a competitive edge . As FHE technology matures and hardware acceleration becomes more prevalent, we can expect to observe wider adoption of privacy-preserving computation across various industries. The theoretical foundations of FHE are well-established, with schemes based on various mathematical problems .