Dataset Viewer
The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.
Couldn't infer the same data file format for all splits. Got {NamedSplit('train'): ('csv', {}), NamedSplit('test'): ('json', {})}
Error code:   FileFormatMismatchBetweenSplitsError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/datasets-cards)

FHE Challenge Benchmark

20 FHE challenges from FHERMA for evaluating AI agents.

Directory Structure

fhe_challenge/
├── black_box/              # Pre-encrypted test cases (3)
│   ├── challenge_relu/
│   ├── challenge_sigmoid/
│   └── challenge_sign/
├── white_box/
│   ├── openfhe/            # OpenFHE C++ challenges (11)
│   ├── ml_inference/       # ML model challenges (4)
│   └── non-OpenFHE/        # Other libraries (2)
└── README.md

Challenge Types

Type Count Key Gen Encrypt Decrypt Agent Task
Black Box 3 Pre-done Pre-done Validator Implement eval()
White Box 17 Validator Validator Validator Implement full solution

Black Box (3 challenges)

Agent receives pre-encrypted inputs and keys. Implements eval() function.

Challenge Scheme Function Input Range Depth
relu CKKS max(0, x) [-1, 1] 12, 4
sigmoid CKKS 1/(1+e^-x) [-25, 25] 7, 4
sign CKKS sign(x) [-1, 1] 10

White Box - OpenFHE (11 challenges)

Challenge Scheme Task
array_sorting CKKS Sort array ascending
gelu CKKS GELU activation function
invertible_matrix CKKS Matrix invertibility check
knn CKKS K-nearest neighbors (k=10)
lookup_table BFV Table lookup by index
matrix_multiplication CKKS 64x64 matrix multiply
max CKKS Find maximum element
parity CKKS Compute parity bit
shl BFV Shift left operation
softmax CKKS Softmax function
svd CKKS SVD decomposition

White Box - ML Inference (4 challenges)

Challenge Model Task Template
cifar10 CNN Image classification (10 classes) C++
house_prediction Linear Price regression Python
sentiment_analysis Classifier Text sentiment (3 classes) Python
svm SVM Fraud detection C++

White Box - Non-OpenFHE (2 challenges)

Challenge Library Scheme Language
string_search IBM HElayers CKKS Python
swift Apple swift-homomorphic-encryption BFV Swift

Running Challenges

Black Box

cd black_box/challenge_relu
docker build -t relu .
docker run --rm -v $(pwd)/tests/testcase1:/data relu

White Box

cd white_box/openfhe/challenge_max
./verify.sh

CLI Arguments

All challenges use consistent CLI args:

  • --cc - CryptoContext
  • --key_pub - Public key
  • --key_mult - Multiplication/relinearization key
  • --key_rot - Rotation keys
  • --output - Output ciphertext path
  • Challenge-specific input args (--input, --sample, --array, etc.)

Resources

Downloads last month
7