Closed tedw-tech closed 3 years ago
Hello, I have tried some tests with v3.4.5, and it's just working fine. I guess there are some mistakes in your evaluator
?
Here is my code
from seal import *
from seal_helper import *
print_example_banner("Example: BFV Basics")
parms = EncryptionParameters(scheme_type.BFV)
poly_modulus_degree = 4096 # you can use yours, it just fast
parms.set_poly_modulus_degree(poly_modulus_degree)
parms.set_coeff_modulus(CoeffModulus.BFVDefault(poly_modulus_degree))
parms.set_plain_modulus(1024) # any positive integer, maybe a prime number and keep it small will be better in theory
context = SEALContext.Create(parms)
print_parameters(context)
evaluator = Evaluator(context)
encoder = IntegerEncoder(context)
keygen = KeyGenerator(context)
public_key = keygen.public_key()
secret_key = keygen.secret_key()
encryptor = Encryptor(context, public_key)
decryptor = Decryptor(context, secret_key)
value1 = -7
plain1 = Plaintext(encoder.encode(value1))
value2 = 25
plain2 = Plaintext(encoder.encode(value2))
x1 = Ciphertext()
encryptor.encrypt(plain1, x1)
evaluator.add_plain_inplace(x1, plain2)
evaluator.multiply_plain_inplace(x1, plain2)
p1 = decryptor.decrypt(x1)
v1 = encoder.decode_int64(p1)
print(v1) # 450
Hey, thanks for the quick reply!
I just ran that code and it does work, however, when I add another multiplication (after the already existing one), the output is 11250
, shouldn't it be 1010
? Or am I confusing the plaintext modulus here with what would happen if I used batching?
(25-7)x25x25=11250, I think it's right. For the detail, you can refer to the notes from the example. or the paper, which implements the BFV scheme.
Ah, sorry, I got something mixed up with the parameters! Never mind, so no problem after all.
I set up the SEAL context with the following values:
params = seal.EncryptionParameters(seal.scheme_type.BFV)
poly_modulus_degree = 16384
params.set_poly_modulus_degree(poly_modulus_degree)
params.set_coeff_modulus(seal.CoeffModulus.BFVDefault(poly_modulus_degree))
plain_mod = 2424833
params.set_plain_modulus(plain_mod)
I am also using the IntegerEncoder, like this:
encoder = seal.IntegerEncoder(context)
value1 = -7
plain1 = seal.Plaintext(encoder.encode(value1))
value2 = 25
plain2 = seal.Plaintext(encoder.encode(value2))
Afterwards, I perform some computations (squarings and multiplications). While the noise budget stays in the double digits, decoding the result to int becomes imposible as I get a Value error.
Shouldn't the result remain in a decoding-range for int64 if it is set to the value above? I also noticed that with smaller input values, I can still decode, but the value is not
mod plaintext_modulus
either. Am I missing something about the purpose of the plain_modulus?