What is the difference between PHE, SHE, and FHE?

Partially Homomorphic Encryption (PHE, e.g. Paillier, RSA) supports only one operation (either addition or multiplication) for an unlimited number of operations. Somewhat Homomorphic Encryption (SHE) supports both addition and multiplication for a limited number of steps before noise corrupts the ciphertext. Fully Homomorphic Encryption (FHE, introduced by Craig Gentry in 2009) supports arbitrary computation for unlimited depth via bootstrapping.

Where is Homomorphic Encryption used in the real world?

Homomorphic Encryption enables privacy-preserving AI inference (evaluating medical diagnoses or credit scores in the cloud without exposing patient data), secure electronic voting systems, encrypted database queries (searching ciphertexts without revealing keywords), and multi-party financial fraud audits.

Homomorphic Encryption Interactive Simulator

Q: What is Homomorphic Encryption?

Homomorphic Encryption is a breakthrough form of cryptography that allows arbitrary mathematical computations to be performed directly on ciphertext without first having to decrypt it. The resulting output remains encrypted, and when decrypted by the private key holder, the plaintext matches the result of operations performed as if on the original unencrypted data.

Compute on encrypted ciphertexts without decrypting. Interactive demonstration of the Paillier Partially Homomorphic Cryptosystem.

🔑 Step 1: Paillier Key Generation

Generated using prime numbers $p = 61$ and $q = 53$ ($n = p \times q = 3233$, $n^2 = 10452289$).

Public Key (n, g):

(3233, 3234)

Private Key (λ, μ):

(780, 1173)

🔒 Step 2: Encrypt Private Numbers (Client-Side)

Private Message 1 ($m_1$):

Ciphertext c₁: Encrypting...

Private Message 2 ($m_2$):

Ciphertext c₂: Encrypting...

☁️ Step 3: Untrusted Cloud Computes on Ciphertext (Zero Knowledge)

The cloud server multiplies the two ciphertexts without possessing the private key:

Homomorphic Addition Property:
$c_{sum} = (c_1 \times c_2) \pmod{n^2}$

Computed Ciphertext c_sum: ...

🎉 Step 4: Client Decrypts with Private Key

Decrypted Result:

Zero Plaintext Leaked to Cloud

What is Homomorphic Encryption?

In traditional cryptography, data must be decrypted into plaintext before it can be processed, searched, or analyzed by a computer. This creates a critical vulnerability: during processing in cloud RAM, sensitive records (medical diagnoses, financial ledger balances, proprietary AI weights) are exposed to memory scrapers, rogue cloud administrators, and subpoena orders.

Homomorphic Encryption (HE) solves this by enabling mathematical operations directly on encrypted ciphertexts. The server computes without ever observing the plaintext, producing an encrypted result that only the private key owner can decrypt.

The Paillier Cryptosystem Mechanics

Invented by Pascal Paillier in 1999, the Paillier cryptosystem is an asymmetric algorithm based on the Decisional Composite Residuosity Assumption. It exhibits additive homomorphic properties:

Homomorphic Addition of Two Plaintexts: Multiplying two ciphertexts modulo $n^2$ results in a ciphertext corresponding to the sum of the plaintexts: $$D(E(m_1, r_1) cdot E(m_2, r_2) pmod{n^2}) = m_1 + m_2 pmod{n}$$
Homomorphic Multiplication by a Constant: Raising a ciphertext to a plaintext scalar $k$ modulo $n^2$ results in a ciphertext corresponding to the product: $$D(E(m_1, r_1)^k pmod{n^2}) = k cdot m_1 pmod{n}$$

Real-World Applications in 2026

Confidential AI Inference: Patients can submit encrypted genomic data to a cloud diagnostic AI model. The model computes on ciphertexts and returns an encrypted diagnosis that only the patient's phone can unlock.
Encrypted Financial Reconciliation: Banks and supply chain partners can cross-audit joint account totals and tax liabilities without revealing individual counterparty sales volumes.
Verifiable Electronic Voting: Tally servers can sum encrypted ballots without decrypting individual voter choices.

Frequently Asked Questions

Can homomorphic encryption perform division or arbitrary logic? +

The Paillier cryptosystem is Partially Homomorphic (PHE) and supports only addition and scalar multiplication. Fully Homomorphic Encryption (FHE schemes like BFV, BGV, and CKKS) supports arbitrary addition and multiplication, enabling complete Turing-complete programs on ciphertexts.

Why isn't all cloud computing encrypted homomorphically today? +

Computational overhead. FHE operations can be 1,000x to 100,000x slower than unencrypted operations due to ciphertext expansion and noise-reduction bootstrapping. However, hardware ASICs and algorithmic breakthroughs are making FHE viable for production financial and healthcare workflows.

Homomorphic Encryption Principles & Zero-Knowledge Computation

Traditional data security models protect information in two primary states: data-at-rest (using AES-256 or ChaCha20 encryption on storage drives) and data-in-transit (using TLS 1.3 across network wires). However, conventional computation requires decrypting ciphertext into unencrypted plaintext in computer memory (RAM), exposing sensitive health records, biometric signatures, and financial transactions to memory-dump vulnerabilities, hypervisor snooping, and rogue server administrators.

1. Classifications of Homomorphic Encryption

Homomorphic Encryption (HE) is a breakthrough cryptographic paradigm that allows mathematical operations to be performed directly on ciphertext, yielding an encrypted result that, when decrypted, matches the output of operations performed on the unencrypted plaintext: Dec(Enc(m1) * Enc(m2)) = m1 + m2.

Partially Homomorphic Encryption (PHE): Supports one specific mathematical operation an unlimited number of times. Prominent examples include RSA (multiplicative homomorphism) and the Paillier Cryptosystem (additive homomorphism).
Somewhat Homomorphic Encryption (SHE): Supports both addition and multiplication, but only for a limited number of sequential operations before accumulated mathematical noise corrupts the ciphertext.
Fully Homomorphic Encryption (FHE): First solved by Craig Gentry in 2009, FHE supports arbitrary, unlimited addition and multiplication operations through recursive 'bootstrapping' noise reduction, allowing arbitrary complex algorithms to execute securely on untrusted third-party clouds.

2. The Paillier Additive Cryptosystem Mechanics

The interactive engine above demonstrates Pascal Paillier's public-key probabilistic asymmetric cryptosystem based on the composite residuosity problem. Paillier encryption satisfies the additive homomorphism property:

The product of two ciphertexts c1 and c2 decrypts to the sum of their corresponding plaintexts m1 + m2:

          D(E(m1) * E(m2) mod n^2) = (m1 + m2) mod n
        

Similarly, raising an encrypted ciphertext to a plaintext scalar power decrypts to the product of the plaintext and scalar: D(E(m1)^k mod n^2) = (k * m1) mod n. This enables confidential banking tallies, encrypted voting counting, and private medical analytics without ever decrypting source datasets.