evalAddCount not working

Library Questions
1

OpenFHE paper says: the user specifies the multiplicative depth (and for some schemes also the maximum number of additions/key-switching operations)
(https://hal-lirmm.ccsd.cnrs.fr/lirmm-04703074v1/document)

And the read-me file at openfhe-development/src/pke/examples/README.md at 6bcca756e9d52b4db3dd2168414df8a7316b1a61 · openfheorg/openfhe-development · GitHub says

**uint32_t evalAddCount (BGV/BFV only)** - maximum number of additions (used for setting noise). In BGV, it is the maximum number of additions at any level.

Based on this, I assume my program (which has 85 cipher additions and 3 cipher multiplications) below should run correctly, but it does not. I am not sure what I am doing wrong here. Please help me.


    CCParams<CryptoContextBGVRNS> parameters;
    parameters.SetMultiplicativeDepth(3);
    parameters.SetEvalAddCount(85);
    parameters.SetPlaintextModulus(536903681);
    parameters.SetMaxRelinSkDeg(4);

    std::vector<int64_t> vectorOfInts1 = {1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
    Plaintext plaintext1               = cryptoContext->MakePackedPlaintext(vectorOfInts1);
    std::vector<int64_t> vectorOfInts2 = {1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1};
    Plaintext plaintext2               = cryptoContext->MakePackedPlaintext(vectorOfInts2);
    std::vector<Ciphertext<DCRTPoly>> ciphertexts;

    ciphertexts.push_back(cryptoContext->Encrypt(keyPair.publicKey, plaintext1));
    ciphertexts.push_back(cryptoContext->Encrypt(keyPair.publicKey, plaintext2));

    auto ciphertextMult12 = cryptoContext->EvalMult(ciphertexts[0], ciphertexts[1]);
    cryptoContext->ModReduceInPlace(ciphertextMult12);

    for(int i=0; i<25; i++){                                                                                                                                                                       
    ciphertextMult12 = cryptoContext->EvalAdd(ciphertextMult12,ciphertextMult12);                                                                                                                     
    }

    Plaintext plaintextDecMult123;
    auto ciphertexts2 = ciphertexts[1];

    for(int i=0; i<=35; i++){                                                                                                                                                                       
    ciphertextMult12 = cryptoContext->EvalAdd(ciphertextMult12,ciphertextMult12);                                                                                                                     
    }

    ciphertextMult12 = cryptoContext->EvalMult(ciphertextMult12, ciphertexts2);
    ciphertextMult12 = cryptoContext->EvalMult(ciphertextMult12, ciphertexts2);

    cryptoContext->RelinearizeInPlace(ciphertextMult12);
    cryptoContext->Decrypt(keyPair.secretKey, ciphertextMult12, &plaintextDecMult123);

    plaintextDecMult123->SetLength(plaintext1->GetLength());
    std::cout << "\nResult of 3 homomorphic multiplications: \n";
    std::cout << plaintextDecMult123 << std::endl;
    for(int i=0; i<=25; i++){
    ciphertextMult12 = cryptoContext->EvalAdd(ciphertextMult12,ciphertexts2);
    }
    cryptoContext->Decrypt(keyPair.secretKey, ciphertextMult12, &plaintextDecMult123);
    plaintextDecMult123->SetLength(plaintext1->GetLength());
    std::cout << plaintextDecMult123 << std::endl

Instead of : ( 86 0 0 0 0 0 0 0 0 0 0 85 ... )

I get
( 118792352 23987527 -207205692 -172374152 3066270 -251675412 183218421 -202333350 46098309 4166024 -89071655 -102998818 ... )

I am not sure what is going on. Is evalAddCount deprecated?
Thank you.

2

The problem you are facing has nothing to do with evalAddCount.
Your computation does not stay within the dynamic range specified in the plaintext modulus.
Your dynamic range is: [-536903681/2,536903681/2] \cap \mathbb{Z}, which is NOT wide enough for your computation in BGV.

What are you trying to do?
Maybe CKKS would be a better choice if you require such high precision.

3

I am not sure why my computation goes outside the dynamic range. I re-run the same program with

    std::vector<int64_t> vectorOfInts1 = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
    std::vector<int64_t> vectorOfInts2 = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};

and got a similar result. If I increased the multiplicative depth, then I have the correct output ( 0 ... ) .

My program involves a combination of additions and multiplication and SetMultiplicativeDepth is not enough for me, I also need to factor in additions while choosing the parameters. I am trying to use EvalAddCount for that purpose.

4

ciphertextMult12 grows exponentially below, no?

Also, you have another for-loop over the same variable.

My understanding is that you are computing the following:

v=1;
for(i=0;i<=24;i++;
   v=v+v;
For(j=0;j<=34;j++;
   v=v+v;

If that is the case, v will be 61-bit long, much larger than your plaintext modulus.

5

Thanks, one thing I failed to see is how you computed 61 bits.

Could you explain? Thank you

6

As follows:

1+1 = 2 = (2^1)
2+2 = 4 = (2^2)
4+4 = 8 = (2^3)
.
.
.
2^59+2^59 = (2^60)

and 2^{60} is a 61-bit integer.