Hi,
I am currently trying to apply the OpenFHE library to my use case:
- Vectors of arbitrary length, either integer or floating point (depending on the applicability of bootstrapping)
- Computation function consisting of multiple multiplications and additions of ciphertexts and plaintexts
- Apply computation function (theoretically) unlimited amount of times to the same resulting ciphertext, without runtime limits
My question is, if it is possible to utilize the OpenFHE library (the python wrapped one in particular) to realize such an application. I tested the “Iterative CKKS bootstrapping” example test and found that the bit precision decreases with each multiplication between two floating point vectors until the point of 0 bit precision is reached and the computation fails.
As far as I understood, using bootstrapping after each multiplication should allow me to apply multiplication to the same ciphertext an unlimited amount of times, if performance and runtime is not an issue.
Is this a wrong assumption? And if I switch to integer vectors, is it then possible to realize arbitrary many computations with OpenFHE?
The result of the computation does not require high precision, it would suffice to provide 2 decimal digits for floating point numbers.
Thanks in advance