Struct seal_fhe::BFVEncoder

pub struct BFVEncoder { /* private fields */ }

Provides functionality for CRT batching. If the polynomial modulus degree is N, and the plaintext modulus is a prime number T such that T is congruent to 1 modulo 2N, then BatchEncoder allows the plaintext elements to be viewed as 2-by-(N/2) matrices of integers modulo T. Homomorphic operations performed on such encrypted matrices are applied coefficient (slot) wise, enabling powerful Batched functionality for computations that are vectorizable. This functionality is often called “batching” in the homomorphic encryption literature.

Mathematical Background

Mathematically speaking, if the polynomial modulus is X^N+1, N is a power of two, and PlainModulus is a prime number T such that 2N divides T-1, then integers modulo T contain a primitive 2N-th root of unity and the polynomial X^N+1 splits into n distinct linear factors as X^N+1 = (X-a_1)*...*(X-a_N) mod T, where the constants a_1, ..., a_n are all the distinct primitive 2N-th roots of unity in integers modulo T. The Chinese Remainder Theorem (CRT) states that the plaintext space Z_T[X]/(X^N+1) in this case is isomorphic (as an algebra) to the N-fold direct product of fields Z_T. The isomorphism is easy to compute explicitly in both directions, which is what this class does. Furthermore, the Galois group of the extension is (Z/2NZ)* ~= Z/2Z x Z/(N/2) whose action on the primitive roots of unity is easy to describe. Since the batching slots correspond 1-to-1 to the primitive roots of unity, applying Galois automorphisms on the plaintext act by permuting the slots. By applying generators of the two cyclic subgroups of the Galois group, we can effectively view the plaintext as a 2-by-(N/2) matrix, and enable cyclic row rotations, and column rotations (row swaps).

Valid Parameters

Whether batching can be used depends on whether the plaintext modulus has been chosen appropriately. Thus, to construct a BatchEncoder the user must provide an instance of SEALContext such that its associated EncryptionParameterQualifiers object has the flags ParametersSet and EnableBatching set to true.

Implementations

impl BFVEncoder

pub fn new(ctx: &Context) -> Result<Self>

Creates a BatchEncoder. It is necessary that the encryption parameters given through the SEALContext object support batching. This means you used PlainModulus::batching when you created your encryption_parameters.

• ctx - The Context

pub fn encode_unsigned(&self, data: &[u64]) -> Result<Plaintext>

Creates a plaintext from a given matrix. This function “batches” a given matrix of integers modulo the plaintext modulus into a plaintext element, and stores the result in the destination parameter. The input vector must have size at most equal to the degree of the polynomial modulus. The first half of the elements represent the first row of the matrix, and the second half represent the second row. The numbers in the matrix can be at most equal to the plaintext modulus for it to represent a valid plaintext.

The matrix’s elements are of type u64.

data - The 2xN matrix of integers modulo plaintext modulus to batch

pub fn encode_signed(&self, data: &[i64]) -> Result<Plaintext>

Creates a plaintext from a given matrix. This function “batches” a given matrix of integers modulo the plaintext modulus into a plaintext element, and stores the result in the destination parameter. The input vector must have size at most equal to the degree of the polynomial modulus. The first half of the elements represent the first row of the matrix, and the second half represent the second row. The numbers in the matrix can be at most equal to the plaintext modulus for it to represent a valid plaintext.

The matrix’s elements are of type i64.

data - The 2xN matrix of integers modulo plaintext modulus to batch

pub fn decode_unsigned(&self, plaintext: &Plaintext) -> Result<Vec<u64>>

Inverse of encode. This function “unbatches” a given plaintext into a matrix of integers modulo the plaintext modulus, and stores the result in the destination parameter. The input plaintext must have degrees less than the polynomial modulus, and coefficients less than the plaintext modulus, i.e. it must be a valid plaintext for the encryption parameters. Dynamic memory allocations in the process are allocated from the memory pool pointed to by the given MemoryPoolHandle.

The input plaintext matrix should be known to contain u64 elements.

• plain - The plaintext polynomial to unbatch

pub fn decode_signed(&self, plaintext: &Plaintext) -> Result<Vec<i64>>

Inverse of encode. This function “unbatches” a given plaintext into a matrix of integers modulo the plaintext modulus, and stores the result in the destination parameter. The input plaintext must have degrees less than the polynomial modulus, and coefficients less than the plaintext modulus, i.e. it must be a valid plaintext for the encryption parameters. Dynamic memory allocations in the process are allocated from the memory pool pointed to by the given MemoryPoolHandle.

The input plaintext matrix should be known to contain i64 elements.

• plain - The plaintext polynomial to unbatch

pub fn get_slot_count(&self) -> usize

Returns the number of “Batched” slots in this encoder produces.

Trait Implementations

impl Drop for BFVEncoder

fn drop(&mut self)

Executes the destructor for this type.

impl Send for BFVEncoder

impl Sync for BFVEncoder