This is the development repository for the OpenFHE library. The current (stable) version is v1.1.4 (released on March 8, 2024).
License: BSD 2-Clause "Simplified" License
Include both single-key and threshold FHE scenarios. Use the method presented in https://eprint.iacr.org/2022/816
cpplint fails while running pre-commit after clang-format passes for styling issues such as "Should have a space between // and comment" errors. This could possibly be fixed using SpacesBeforeTrailingComments from https://clang.llvm.org/docs/ClangFormatStyleOptions.html automatically with clang-format instead of fixing them manually.
Keyswitching documentation is stale for keyswitch-bv.h and keyswitch-hybrid.h. They seem to reference the documentation for the base class?
Can someone confirm if the documentation of keyswitch-BV.cpp and keyswitch-hybrid.cpp and keyswitch-rns.cppare correct? They all have the same documentation discussing CKKS?
/*
CKKS scheme implementation
*/
/*
Description:
This code implements RNS variants of the Cheon-Kim-Kim-Song scheme.
The CKKS scheme is introduced in the following paper:
- Jung Hee Cheon, Andrey Kim, Miran Kim, and Yongsoo Song. Homomorphic
encryption for arithmetic of approximate numbers. Cryptology ePrint Archive,
Report 2016/421, 2016. https://eprint.iacr.org/2016/421.
Our implementation builds from the designs here:
- Marcelo Blatt, Alexander Gusev, Yuriy Polyakov, Kurt Rohloff, and Vinod
Vaikuntanathan. Optimized homomorphic encryption solution for secure genomewide
association studies. Cryptology ePrint Archive, Report 2019/223, 2019.
https://eprint.iacr.org/2019/223.
- Andrey Kim, Antonis Papadimitriou, and Yuriy Polyakov. Approximate
homomorphic encryption with reduced approximation error. Cryptology ePrint
Archive, Report 2020/1118, 2020. https://eprint.iacr.org/2020/
1118.
*/
keyswitch-bv and keyswitch-hybrid? I don't know enough to comment on them or write documentation up.See openfhe discussion: Open FHE Building Error in Windows 10 for more details.
Remove exra parameters and normalize the names, like depth and maxDepth
GetParamsPK() uses extended basis for the HYBRID mode. This is needed to support the PRE functionality. Currently this extended basis mode is used for all HYBRID scenarios, not just those involving PRE.
Once the PRE changes by @sararv22 are merged, the PRE mode data member can be used to restrict the use of the extended basis to HYBRID + PRE scenarios only.
The use of extended basis increases the KeyGen runtime and slightly increases the public key encryption for all cases when the HYBRID key switching method is used.
The changes for the implementation of GetParamsPK() need to be made rlw-cryptoparameters.h.
When I want to serialize the EvalMultKey or EvalAutomorphismKey exactly like in "simple-real-numbers-serial.cpp" I get this error :
libc++abi: terminating with uncaught exception of type cereal::Exception: Trying to save an unregistered polymorphic type (lbcrypto::EvalKeyRelinImpl<lbcrypto::DCRTPolyImpl<bigintdyn::mubintvec<bigintdyn::ubint > > >).
Make sure your type is registered with CEREAL_REGISTER_TYPE and that the archive you are using was included (and registered with CEREAL_REGISTER_ARCHIVE) prior to calling CEREAL_REGISTER_TYPE.
If your type is already registered and you still see this error, you may need to use CEREAL_REGISTER_DYNAMIC_INIT.
example: MakeCKKSPackedPlaintext() in src/pke/include/cryptocontext.h
This may become an issue as the overriders of virtual functions do not acquire the default arguments from the base class declarations, and when the virtual function call is made, the default arguments are decided based on the static type of the object.
I have not been able to replicate the issue on my machine, but here is an example run where it seg faults: https://github.com/openfheorg/openfhe-development/runs/7136560551?check_suite_focus=true
HYBRID key switching is implemented for the parent RNS scheme. We will need some special logic for BFV to support this.
seems to explicitly use a variable after freeing it. I'm not a C++ expert, but this is causing build failures in GCC 12.
Repro steps: clone this repo and follow the instructions in the Makefile
build git:(main) $ make
-- Copied demoData files
[ 0%] Built target third-party
[ 0%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/be2-poly-impl.cpp.o
[ 0%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/be4-poly-impl.cpp.o
[ 1%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/be6-poly-impl.cpp.o
[ 1%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/benative-poly-impl.cpp.o
[ 1%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/dgsampling.cpp.o
[ 2%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/elemparamfactory.cpp.o
[ 2%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/elemparams.cpp.o
[ 2%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/field2n.cpp.o
[ 3%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/hal/default/dcrtpoly-impl.cpp.o
In file included from /home/jkun/openfhe-development/src/core/include/math/discretegaussiangenerator.h:77,
from /home/jkun/openfhe-development/src/core/include/lattice/ilelement.h:42,
from /home/jkun/openfhe-development/src/core/include/lattice/hal/dcrtpoly-interface.h:52,
from /home/jkun/openfhe-development/src/core/include/lattice/hal/default/dcrtpoly.h:51,
from /home/jkun/openfhe-development/src/core/include/lattice/hal/default/lat-backend-default.h:39,
from /home/jkun/openfhe-development/src/core/include/lattice/lat-hal.h:42,
from /home/jkun/openfhe-development/src/core/lib/lattice/hal/default/dcrtpoly.cpp:39,
from /home/jkun/openfhe-development/src/core/lib/lattice/hal/default/dcrtpoly-impl.cpp:36:
/home/jkun/openfhe-development/src/core/include/math/distributiongenerator.h: In static member function ‘static lbcrypto::PRNG& lbcrypto::PseudoRandomNumberGenerator::GetPRNG()’:
/home/jkun/openfhe-development/src/core/include/math/distributiongenerator.h:132:26: error: pointer ‘mem’ used after ‘void free(void*)’ [-Werror=use-after-free]
132 | uint32_t counter = reinterpret_cast<long long>(mem); // NOLINT
| ^~~~~~~
/home/jkun/openfhe-development/src/core/include/math/distributiongenerator.h:131:21: note: call to ‘void free(void*)’ here
131 | free(mem);
| ~~~~^~~~~
cc1plus: all warnings being treated as errors
make[2]: *** [src/core/CMakeFiles/coreobj.dir/build.make:188: src/core/CMakeFiles/coreobj.dir/lib/lattice/hal/default/dcrtpoly-impl.cpp.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:551: src/core/CMakeFiles/coreobj.dir/all] Error 2
make: *** [Makefile:136: all] Error 2
The following error is generated:
[ 12%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/trapdoor-poly-impl.cpp.o
error: use of undeclared identifier 'Mul128'
sum[j] += Mul128(a, b);
^
[ 13%] Building CXX object src/core/CMakeFiles/coreobj.dir/lib/lattice/trapdoor.cpp.o
In file included from /home/ypolyakov/git/openfhe-development/src/core/lib/lattice/hal/default/dcrtpoly-impl.cpp:36:
/home/ypolyakov/git/openfhe-development/src/core/lib/lattice/hal/default/dcrtpoly.cpp:1757:27: error: use of undeclared identifier 'Mul128'
sum[j] += Mul128(a, b);
^
1 error generated.
make[2]: *** [src/core/CMakeFiles/coreobj.dir/build.make:180: src/core/CMakeFiles/coreobj.dir/lib/lattice/hal/default/dcrtpoly.cpp.o] Error 1
make[2]: *** Waiting for unfinished jobs....
/home/ypolyakov/git/openfhe-development/src/core/lib/lattice/hal/default/dcrtpoly.cpp:1763:39: error: use of undeclared identifier 'BarrettUint128ModUint64'
partPl.m_vectors[j][ri] = BarrettUint128ModUint64(sum[j], pj.ConvertToInt(), precomputed.modpBarrettMu[j]);
^
/home/ypolyakov/git/openfhe-development/src/core/lib/lattice/hal/default/dcrtpoly-impl.cpp:50:16: note: in instantiation of member function 'lbcrypto::DCRTPolyImpl<bigintfxd::BigVectorFixedT<bigintfxd::BigIntegerFixedT<unsigned int, 3500> > >::FastExpandCRTBasisPloverQ' requested here
template class DCRTPolyImpl<M2Vector>;
^
In file included from /home/ypolyakov/git/openfhe-development/src/core/lib/lattice/hal/default/dcrtpoly-impl.cpp:36:
/home/ypolyakov/git/openfhe-development/src/core/lib/lattice/hal/default/dcrtpoly.cpp:1763:39: error: use of undeclared identifier 'BarrettUint128ModUint64'
partPl.m_vectors[j][ri] = BarrettUint128ModUint64(sum[j], pj.ConvertToInt(), precomputed.modpBarrettMu[j]);
^
/home/ypolyakov/git/openfhe-development/src/core/lib/lattice/hal/default/dcrtpoly-impl.cpp:52:16: note: in instantiation of member function 'lbcrypto::DCRTPolyImpl<bigintdyn::mubintvec<bigintdyn::ubint<unsigned int> > >::FastExpandCRTBasisPloverQ' requested here
template class DCRTPolyImpl<M4Vector>;
^
3 errors generated.
make[2]: *** [src/core/CMakeFiles/coreobj.dir/build.make:167: src/core/CMakeFiles/coreobj.dir/lib/lattice/hal/default/dcrtpoly-impl.cpp.o] Error 1
[ 13%] Linking CXX static library ../../../lib/libbenchmark.a
[ 13%] Built target benchmark
make[1]: *** [CMakeFiles/Makefile2:571: src/core/CMakeFiles/coreobj.dir/all] Error 2
make: *** [Makefile:130: all] Error 2
Compiler error for 128-bit link on MacOS:
src/pke/unittest/utckksrns/UnitTestCKKSrns.cpp:160:17: error: unused variable 'RING_DIM_PREC'
When the exception in DCRTPolyImpl::Times is enabled, it gets triggered during CKKS bootstrapping. See example below. Need to fix the root cause and reenable the exception.
bin/examples/pke/ckks-bootstrap
SecretKeyDist: UNIFORM_TERNARY
p = 59
n = 4096
# slots = 2048
log2 q = 1948
log2 q*p = 2608
dim1 = [ 0 0 ]
level budget = [ 4 4 ]
Linear transform precomputation time: 1.72 s
Automorphism key generation time: 2.175 s
ciphertext number of slots: 2048
Number of levels before bootstrapping: 1
terminate called after throwing an instance of 'lbcrypto::math_error'
what(): /home/ypolyakov/git/openfhe-development/src/core/lib/lattice/hal/default/dcrtpoly.cpp:914 tower size mismatch; cannot multiply
Aborted (core dumped)
Some MULTIHOP_PRE_TESTs fail at times with the new moduli estimates in FixedAuto. The moduli estimates are located in bgvrns-parametergeneration.cpp. As a workaround, we currently estimate the keySwitchingNoise to use numPrimes + 1, resulting in larger moduli than necessary. This test fails when we use numPrimes.
in pke/lib/scheme/bfvrns/bfvrns-leveledshe.cpp
In LeveledSHECKKSRNS class several operations: ciphertext +/-/* plaintext are implemented using workaround.
First we decode the plaintext and then encode again with the proper depth/level.
Instead, we can treat a plaintext as a "dummy" ciphertext, and do all operations in a same way as ciphertext * ciphertext.
P.S. we also can redefine the Plaintext to smth like PlainCiphertext. And define a Plaintext to be not depend on depth/level, and instead Plaintext will be
example: LeveledSHEBFVRNS::EvalMult() in src/pke/lib/scheme/bfvrns/bfvrns-leveledshe.cpp
the return type of GetCryptoParameters() is const std::shared_ptr<CryptoParametersBase>. It is cast to CryptoParametersBFVRNS using std::static_pointer_cast<>
Currently they are documented only in the code: https://github.com/openfheorg/openfhe-development/blob/main/src/pke/include/scheme/cryptocontextparams-base.h
Some headers include other files unnecessarily. Some of them are:
scheme/bgvrns/bgvrns-scheme.h
key/allkey.h
Some headers include too many header files. Example:
scheme/allscheme.h
Currently the parameters for approximating the modular reduction using the sine wave in the 64-bit scenario are tailored for FLEXIBLEAUTO. A better precision can be achieved for FLEXIBLEAUTOEXT if a different (larger) ratio of q/m is used.
In the EncodeVec function of the CoefPackedEncoding, when encoding negative values, we encode t - abs(value), where t is the plaintext modulus. Before, we used q - abs(value), where q is the ciphertext modulus because this has less noise. Why does using q fail with the noise improvement when using Q/t instead of Delta?
The following two optimizations can be applied
See the example below
bin/examples/binfhe/eval-flooring
Generating the bootstrapping keys...
Completed the key generation.
Homomorphically round down the input by 2 bits.
Input: 12. Expected: 3. Evaluated = 1
http://homomorphicencryption.org/wp-content/uploads/2018/11/HomomorphicEncryptionStandardv1.1.pdf does not include entries for
Some unit tests are currently commented out because they are failing
Even by following the instructions for macOS, openfhe fails to compile on macOS (M2 chip). See below for the error that seems due to a deprecated member.
[ 20%] Building CXX object src/core/CMakeFiles/core_tests.dir/unittest/UnitTestSerialize.cpp.o
In file included from /Users/tancrede/git/openfhe-development/src/core/unittest/UnitTestSerialize.cpp:49:
In file included from /Users/tancrede/git/openfhe-development/src/core/include/utils/serial.h:67:
In file included from /Users/tancrede/git/openfhe-development/third-party/cereal/include/cereal/archives/json.hpp:62:
/Users/tancrede/git/openfhe-development/third-party/cereal/include/cereal/external/rapidjson/document.h:102:19: error: 'iterator<std::random_access_iterator_tag, rapidjson::GenericMember<rapidjson::UTF8<>, rapidjson::MemoryPoolAllocator<>>>' is deprecated [-Werror,-Wdeprecated-declarations]
: public std::iterator<std::random_access_iterator_tag
^
/Users/tancrede/git/openfhe-development/third-party/cereal/include/cereal/external/rapidjson/document.h:548:22: note: in instantiation of template class 'rapidjson::GenericMemberIterator<false, rapidjson::UTF8<>, rapidjson::MemoryPoolAllocator<>>' requested here
typedef typename GenericMemberIterator<false,Encoding,Allocator>::Iterator MemberIterator; //!< Member iterator for iterating in object.
^
/Users/tancrede/git/openfhe-development/third-party/cereal/include/cereal/archives/json.hpp:440:15: note: in instantiation of template class 'rapidjson::GenericValue<rapidjson::UTF8<>>' requested here
typedef JSONValue::ConstMemberIterator MemberIterator;
^
/Library/Developer/CommandLineTools/SDKs/MacOSX12.3.sdk/usr/include/c++/v1/__iterator/iterator.h:27:29: note: 'iterator<std::random_access_iterator_tag, rapidjson::GenericMember<rapidjson::UTF8<>, rapidjson::MemoryPoolAllocator<>>>' has been explicitly marked deprecated here
struct _LIBCPP_TEMPLATE_VIS _LIBCPP_DEPRECATED_IN_CXX17 iterator
^
/Library/Developer/CommandLineTools/SDKs/MacOSX12.3.sdk/usr/include/c++/v1/__config:1066:39: note: expanded from macro '_LIBCPP_DEPRECATED_IN_CXX17'
# define _LIBCPP_DEPRECATED_IN_CXX17 _LIBCPP_DEPRECATED
^
/Library/Developer/CommandLineTools/SDKs/MacOSX12.3.sdk/usr/include/c++/v1/__config:1043:48: note: expanded from macro '_LIBCPP_DEPRECATED'
# define _LIBCPP_DEPRECATED __attribute__ ((deprecated))
^
The output:
===== HybridKeySwitchingDemo1 =============
CKKS scheme is using ring dimension 32768
Input x: (1, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, ... ); Estimated precision: 90 bits
Segmentation fault (core dumped)
Currently we add a Gaussian with a standard deviation of
The EvalMultMany function returns the incorrect answer sometimes. For example, when the multiplicative depth is 4, and we input 2^4 = 16 ciphertexts, this returns an incorrect answer for BFV modes HPS, HPSPOVERQ, and HPSPOVERQLEVELED. However, it returns the correct answer when we manually multiple 16 ciphertexts with EvalMult.
Looks like we still use cereal from https://gitlab.com/palisade/cereal which was updated the last time in Jan 2020. If we do then either we should get the cereal code directly from https://github.com/USCiLab/cereal or our repo with cereal should be updated with the latest version of cereal.
See https://eprint.iacr.org/2021/1337 for details
The macro defined in this header is not used anymore.
Dockerfile has an old git path, branch and tag. It worked for me this way:
Merge evalSumKeyMap, evalAutomorphismKeyMap
Need a logic on which keys should be serialized/deserialized depending on which functionalities we want to support
Need flags on which functionalities currently are supported
If some keys are missing, we can use other keys to evaluate the desired rotation.
If we construct a BinFHEContext and LWEPrivateKey pair:
cc.GenerateBinFHEContext(kSecurityLevel);
sk = cc.KeyGen();
cc.BTKeyGen(sk);
then attempt to clear them both using:
cc.ClearBTKeys();
sk.reset();
and then re-generate the pair, we seem to get corrupt results for all BinFHE operations.
To reproduce: install bazelisk, unzip openfhe-bug.zip, and then run bazelisk test -c opt //:cleanup_and_run_test inside the resulting folder.
Alternatively, the actual test is a single file (cleanup_and_run_test.cc), so any build system that can handle a dependency on gtest and OpenFHE should work.
The following bugfixes from v1.11.7 (https://gitlab.com/palisade/palisade-development/-/milestones/22#tab-issues) also need to be applied to OpenFHE:
In the bgv-bfv-benchmarks branch, if we run the bgv-bfv-benchmarks for the Polynomial demo, there are some incorrect results for degree 48 and 64 polynomials for HPS and BEHZ. It looks like the noise level reaches 58, while the modulus is 60. Maybe there is a bug in BFV moduli selection in ParamsGen.
Review the use of depth
Review the use of level
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.