privacy-preserving-ml

A curated list of awesome open source libraries to deploy, monitor, version and scale your machine learning

An Open Framework for Federated Learning.

Synthetic benchmark for privacy-preserving and fairness-aware ranking under signal loss

Distributed peer-to-peer LLM inference. Your prompt never leaves your device in clear text.

Official code for "DCT-CryptoNets: Scaling Private Inference in the Frequency Domain" [ICLR 2025]

Secure Federated Learning system with Byzantine attack detection, trust scoring, and real-time SOC dashboard. Built with Flower (flwr), PyTorch, FastAPI, and Next.js. Final Year Project — Bahria University 2026.

A Privacy-Preserving Federated Learnig benchamarking framework, based on TensorFlow/Keras and OpenFHE

A curated collection of privacy-preserving machine learning techniques, tools, and practical evaluations. Focuses on differential privacy, federated learning, secure computation, and synthetic data generation for implementing privacy in ML workflows.

This repository explores federated deep generative models with PyTorch, featuring Conditional DCGAN, FedGAN v2, and custom synchronization strategies. It demonstrates client-server training with FedAvg, non-IID data splits, and GAN evaluation, providing a foundation for research in privacy-preserving generative modeling.

Repo for Mphasis PPML Research Project

Automates hermetic environments (macOS/HPC) to eliminate drift. Provisions offline RAG (Gemma 2), compiles LaTeX manuscripts, and indexes local knowledge. Unifies infrastructure, writing, and inference into a single, audit-ready artifact.

DP-SGD accuracy vs. privacy budget (ε) experiments on MNIST using Opacus — quantifies the accuracy/privacy/compute trade-off with RDP accounting.

Reference implementation of the BHDR regression kernel: BSGS-hoisted diagonal Kernel SHAP regression under CKKS FHE.

Research prototype for real Mamba-2-130M inference under CKKS/FHE (OpenFHE/FIDESlib-GPU): +0.12% PPL surrogate and a verified 24-layer, 3-token encrypted B300 path.

Federated learning + iDLG gradient inversion attack + Central DP defense + Gradio demo. The honest finding: naive Central DP collapses utility on small federations (Gaussian-mechanism curse of dimensionality). Production fixes (DP-SGD, Opacus) documented.

Sovereign Map is a production-grade, Byzantine-tolerant Federated Learning framework. Utilizing the Mohawk Protocol for streaming aggregation, it achieves a 224x memory reduction, enabling secure orchestration of 100M+ nodes via TPM 2.0 hardware-rooted trust. Features full-stack observability with Prometheus & Grafana, built-in tokenomics telemetry

A minimal, hardened Rust runtime for executing critical federated learning aggregation logic (e.g., Multi-Krum, Federated Averaging) entirely within hardware-enforced Trusted Execution Environments (TEEs) including AMD SEV-SNP, Intel SGX/TDX, and AWS Nitro Enclaves.

SecureMed-LLM: A privacy-preserving framework for clinical report generation from chest X-rays, integrating Med-Guard anonymization, DP-SGD (ε=3.0), adversarial training, IDS-LLM validation, and ECIES/Curve25519 encryption. PeerJ Computer Science 2025.

TZDC - A Python library for privacy-enhancing data operations using cryptographic fragmentation and temporal key expiration.

Banks jointly train a fraud/AML model to catch cross-institutional laundering rings none can see alone - without sharing raw data, with differential privacy bounding leakage (explicit ε).