Problems in "OpenFHE integer examples" project

Library Questions
,
1

I’m having weird behaviour when running openfhe-integer-examples project (both examples).
More details here: mismatch in the total occurences between the cleartext and encrypted versions · Issue #2 · openfheorg/openfhe-integer-examples · GitHub

Does anybody else get the same behavior?

2

I’m running into plaintext extra finds X Y so confirmed

3

It seems incorect resuts occur when setting the multiplicative depth too high.

  1. For strsearch_enc_1, setting the depth less than 18 returns correct results.
  2. For strsearch_enc_2 with SIMD, setting the depth low enougth while adjusting the text size and the maximum number of batches also returns correct results.

This is counterintuitive, how setting a higher depth than required causes problems?

4

The problem is now fixed (in the master branch of GitHub - openfheorg/openfhe-integer-examples: Integer examples for OpenFHE library. Based off of https://gitlab.com/palisade/palisade-integer-examples). Please try it when you get a chance.

2 Likes
5

I see that the fix was in changing the RNS variants of the BFV scheme.
@ypolyakov can you maybe give a guideline on when should I use each variant?

6

All RNS variants should work correctly and have roughly the same noise requirements. It appears we have a bug with HPSPOVERQ and HPSPOVERQLEVELED in OpenFHE. I created an issue for it: Bug in HPSPOVERQ and HPSOVERQLEVELED · Issue #280 · openfheorg/openfhe-development · GitHub