Privacy-preserving LLM inference with CKKS homomorphic encryption and Private Linear Layer (PLL) protection for LoRA fine-tuned models
-
Updated
Oct 18, 2025 - Python
Privacy-preserving LLM inference with CKKS homomorphic encryption and Private Linear Layer (PLL) protection for LoRA fine-tuned models
Fully Homomorphic Encryption for Private Federated Learning
Depth-tracked regulatory audit primitives for privacy-preserving AI audits with signed envelopes and TenSEAL CKKS support.
A scalable, Fully Homomorphic Encryption (FHE) pipeline that allows for model inference on encrypted data without the need for decryption.
Drop-in encrypted Fairlearn metrics over CKKS. Same API surface; ciphertext arithmetic via TenSEAL or OpenFHE.
Collaborative Machine Learning approach to train a mode that classifies a person as smoker or non-smoker based on the user data. The distributed approach of training is done with secure model transmissions to central cloud location where Amazon EC2 instance aggregates the new model based on new training received in Homomorphically Encrypted forms
A Two-Party Secure Computation Protocol using Homomorphic Encryption
This repository contains the practical component developed for my Applied Research Project CA 1 at Dublin Business School - MSc in Information Systems with Computing, Year 1, Semester 3.
PrivAnalytica is a prototype Encrypted Analytics-as-a-Service platform that uses TenSEAL (CKKS) to encrypt numeric datasets so statistical computations and ML inference can be performed on ciphertexts and server never sees. Results are returned in encrypted form and require a secret key to decrypt, preserving data confidentiality during processing.
Using fully homomorphic encryption (TenSEAL) for encryption and storage of biometric data.
Adaptive Noise-Aware Neural Network Training on CKKS-Encrypted Data. Trains a neural net entirely on ciphertexts (TenSEAL/Microsoft SEAL); an adaptive capacity-aware controller cuts ciphertext refreshes by 55% and training time by 42% versus a tuned fixed-schedule baseline, at identical accuracy.
Privacy-preserving biometric ML with PySyft + TenSEAL (homomorphic encryption / federated training)
Encrypted memory and search for AI apps.
This repository contains the practical component developed for my Applied Research Project CA 1 at Dublin Business School - MSc in Information Systems with Computing, Year 1, Semester 3.
Privacy-preserving disease risk prediction using the CKKS homomorphic encryption scheme.
Federated learning with homomorphic-encrypted gradient aggregation — the server can compute on encrypted gradients but can never decrypt them.
Privacy-preserving IoT data transaction system combining blockchain, IPFS, access control, encrypted computation, and ML-driven pricing.
Reproducible benchmark of Homomorphic Encryption (CKKS/TenSEAL) vs AES-256 and RSA-2048 for numerical analytics: runtime, memory, ciphertext size and approximation error. Confidential Computing course project.
Secure Cricket RAG Intelligence Engine — RAG pipeline with CKKS Fully Homomorphic Encryption | Llama 3.1/3.2 + ChromaDB + TenSEAL | Airtel Summer Internship 2026 | IIITDM Kurnool
To associate your repository with the tenseal topic, visit your repo's landing page and select "manage topics."