Privacy Paradox Breakthrough: Open-Source LLMs as Data Generators for Secure IoT Infrastructure

This is an official implementation of the following paper:

Youngjoon Lee, Jinu Gong, Taehyun Park,and Joonhyuk Kang. Privacy Paradox Breakthrough: Open-Source LLMs as Data Generators for Secure IoT Infrastructure
submitted.

Requirements

The synthetic data generation (synthesize.py) implementation runs on

docker pull ollama/ollama

The training and test (main.py) implementation runs on

docker pull vault.habana.ai/gaudi-docker/1.20.0/ubuntu22.04/habanalabs/pytorch-installer-2.6.0:latest

Additionally, please install the required packages as below

pip install ollama transformers pandas numpy scikit-learn tqdm

Dataset

• Medical Abstracts Text Classification Dataset (Schopf, Tim, et al. ACM NLPIR 2017)