kyoohyunghan / fullrns-heaan Goto Github PK

About security level of ciphertexts in HEAAN (or any FHE scheme for that matter)
If we see sagemathcell implementation of tester for checking security levels for given paramters, as we increase logq for same n value, it says security level also increases. is this true??
If so why not just set n to 1024 and increase only logq to meet different application requirements, i.e. higher multiplicative depth??

Hello,

I've found some cases in which the homomorphic multiplication seems to fail. In particular, I got this MWE that tries to compute -22.5 * -22.5:

	long logN = 7;
	long L = 2;
	long logp = 45;
	long logSlots = 1;

	long K = L + 1;
	Context context(logN, logp, L, K);
	SecretKey secretKey(context);
	Scheme scheme(secretKey, context);

	srand(0);

	complex<double> mvec1 = {-22.5,0};
	complex<double> mvec2 = {-22.5,0};

	complex<double> mvecMult;

	mvecMult = mvec1 * mvec2;

	Ciphertext cipher1 = scheme.encryptSingle(mvec1, L);
	Ciphertext cipher2 = scheme.encryptSingle(mvec2, L);

	Ciphertext multCipher = scheme.mult(cipher1, cipher2);
	scheme.reScaleByAndEqual(multCipher, 1);

	complex<double>* dvecMult = scheme.decrypt(secretKey, multCipher);

        std::cout << "Expected: " << (mvec1 * mvec2) << std::endl;
        std::cout << "Output: " << *dvecMult << std::endl;

This is the output I get:

Expected: (506.25,-0)
Output: (-18073.4,-30416)

Still, it works for 22.5 * 22.5.
Any ideas about what is happening here?

Do you have plans to increase the code of the bootstrapping part?

According to your article "A Full RNS Variant of Approximate Homomorphic Encryption", the Vector qVec should have L + 1 elements. However, it seems like that the size of qVec may be L or L + 1 as shown in the code. I wonder if there is something wrong?

I am sorry to trouble you.
When i use FullRNS-HEAAN to do a negative number multiplication, the answer is wrong.
Can FullRNS-HEAAN do negative number multiplication?(negative number addition is right)
Thanks a million.