PODT is a CMake-based C++ research artifact for secure decision tree evaluation. This repository accompanies the following paper:

J. Li and L. F. Zhang. Silent Guardians: Independent and Secure Decision Tree Evaluation Without Chatter. in IEEE Transactions on Dependable and Secure Computing, vol. 23, no. 4, pp. 7691-7708, 2026. https://doi.org/10.1109/TDSC.2026.3675596

The repository contains three protocol implementations:

This code is intended for experimental evaluation and reproducibility. It is not a production cryptographic library.

.
├── CMakeLists.txt
├── CMakePresets.json
├── cmake/
├── common/
├── ODTE_SH/
├── PVODTE_MS/
├── PVODTE_MH/
└── .github/workflows/ci.yml

• CMake >= 3.25
• A C++11-compatible compiler
• NTL
• GMP

Example package names:

# Ubuntu / Debian
sudo apt-get update
sudo apt-get install -y cmake g++ libntl-dev libgmp-dev

# macOS with Homebrew
brew install cmake ntl gmp

Configure and build:

cmake --preset default
cmake --build --preset default

Run the repository smoke tests:

ctest --preset default

Run one protocol directly:

./build/default/ODTE_SH/ODTE_SH --quick
./build/default/PVODTE_MS/PVODTE_MS --quick
./build/default/PVODTE_MH/PVODTE_MH --quick

If the dependencies are not in a standard search path, configure with an explicit prefix:

cmake -S . -B build -DCMAKE_PREFIX_PATH=/opt/homebrew

You can also pass NTL_DIR and GMP_DIR directly:

cmake -S . -B build -DNTL_DIR=/path/to/ntl -DGMP_DIR=/path/to/gmp

The default commands are:

./build/default/ODTE_SH/ODTE_SH
./build/default/PVODTE_MS/PVODTE_MS
./build/default/PVODTE_MH/PVODTE_MH

These run the repository's default workload family:

• tree depths: 3, 8, 13, 17, 20
• attribute counts: 13, 9, 13, 57, 784
• message bit-length: 10
• repetitions: 5

Use --quick only for a fast end-to-end check. It runs a reduced case and is not intended for paper timings.

All three executables share the same CLI:

./build/default/ODTE_SH/ODTE_SH --help
./build/default/ODTE_SH/ODTE_SH --depth 3 --attributes 13 --msgbit 10 --cycles 3

Supported options:

• --quick: run a reduced smoke-test case
• --depth <n>: set tree depth for a single custom run
• --attributes <n>: set attribute count for a single custom run
• --msgbit <n>: set message bit-length
• --cycles <n>: set timing repetitions
• --debug: enable debug mode
• --help: print usage

To compare all three protocols on the same parameter point:

./build/default/ODTE_SH/ODTE_SH --depth 3 --attributes 13 --msgbit 10 --cycles 3
./build/default/PVODTE_MS/PVODTE_MS --depth 3 --attributes 13 --msgbit 10 --cycles 3
./build/default/PVODTE_MH/PVODTE_MH --depth 3 --attributes 13 --msgbit 10 --cycles 3

Each run prints phase-wise timing statistics such as:

• Setup algo time
• Provider encryption time
• Feature Selection time
• HSSCMP or VHSSCMP time
• ClassificationGen time
• DTevaluation 0/1 time
• Decryption time

The reported values are mean time in milliseconds plus RSD (relative standard deviation). In --quick mode the default repetition count is 1, so RSD is usually 0%.

• ctest --preset default currently runs the same --quick mode for all three executables.
• The benchmark drivers generate inputs internally; this repository is aimed at protocol evaluation, not deployment.
• The repo includes CMakePresets.json, .editorconfig, .clang-format, and GitHub Actions CI for basic project hygiene.

If you use this repository in academic work, please cite:

@article{li2026silent,
  author = {J. Li and L. F. Zhang},
  title = {Silent Guardians: Independent and Secure Decision Tree Evaluation Without Chatter},
  journal = {IEEE Transactions on Dependable and Secure Computing},
  year = {2026},
  volume = {23},
  number = {4},
  pages = {7691--7708},
  doi = {10.1109/TDSC.2026.3675596}
}

Released under the MIT License. See LICENSE for details.