Skip to main content
Definition Website Computer and Technology

Homomorphic Encryption

Subject: Computer and Technology noun /ˌhoʊ.məˈmɔːr.fɪk ɪnˈkrɪp.ʃən/ 1 definition

  1. Definition (noun)

    Entry #192 · Updated

    Homomorphic Encryption (HE) is a cryptographic technique that preserves algebraic structure between plaintext and ciphertext, allowing an untrusted party to evaluate functions directly on encrypted data without access to the secret decryption key. When the resulting ciphertext is decrypted by the keyholder, the output matches the result of applying those same operations to the unencrypted inputs. Unlike conventional encryption—which protects data only at rest or in transit because it must be decrypted before processing—homomorphic encryption maintains confidentiality during active computation, trading higher computational overhead and larger ciphertext sizes for end-to-end processing privacy.

    Common Types

    • Partially Homomorphic Encryption (PHE): Supports an unlimited number of evaluations for only a single algebraic operation—either addition (such as Paillier) or multiplication (such as textbook RSA or ElGamal)—keeping overhead practical for narrow workloads like encrypted tallies.

    • Somewhat Homomorphic Encryption (SHE): Evaluates circuits combining both addition and multiplication, but only up to a small, fixed complexity because cryptographic noise accumulates with each operation until decryption fails.

    • Leveled Fully Homomorphic Encryption (Leveled FHE): Supports both operations up to a predetermined multiplicative depth chosen during setup, sizing encryption parameters to accommodate the circuit's noise budget without invoking costly ciphertext-refresh steps.

    • Fully Homomorphic Encryption (FHE): Enables arbitrary, unbounded sequences of addition and multiplication by periodically running a bootstrapping procedure that evaluates the decryption circuit homomorphically to reset accumulated noise.

    Example

    A hospital encrypts patient biomarker vectors using a CKKS homomorphic encryption scheme before uploading them to an untrusted cloud provider to run a cardiovascular risk-scoring model. The cloud server evaluates the model's weighted additions and multiplications directly on the ciphertexts and returns an encrypted risk score that only the hospital can decrypt with its secret key.