Enabling Computation on Encrypted Data

In 2009, researchers introduced Fully Homomorphic Encryption (FHE), solving a decades-old open problem in cryptography: how to compute directly on encrypted data. Traditional encryption requires information to be decrypted before processing, a moment when it becomes vulnerable. FHE changed that by allowing computation to happen while the data remains encrypted from start to finish.

What makes this breakthrough revolutionary is that FHE is Turing-complete: in theory, any computation that can be done on plaintext can also be done on ciphertext. It’s not limited to simple operations and can extend across all data types and programs, from basic arithmetic to complex AI models. In essence, this is what we call encrypted compute: performing computation entirely within the encrypted domain so that data is never exposed.

Today, FHE is one of the fastest-growing fields in privacy-preserving technology, with NIST actively working to establish open standards for secure and interoperable implementations. While the potential is transformative, FHE computations are still significantly slower than their plaintext equivalents — a limitation that has historically kept it in the research domain. But with advances in algorithms, hardware acceleration, and dedicated chip architectures, this is rapidly changing.

Before FHE, the main approach to processing sensitive data securely relied on Trusted Execution Environments (TEEs), specialized hardware enclaves where data is decrypted and processed inside a protected space. While TEEs marked an important step forward, they still depend on trusting the underlying hardware, firmware, and operators, a trust that can be broken through hardware vulnerabilities or insider access.

FHE removes that dependency altogether. By replacing trust in hardware with trust in mathematics, it enables true end-to-end protection, turning data privacy from a policy choice into a mathematical guarantee.