Attacks Against ML Models
Techniques for fooling or bypassing machine-learning models. Covers adversarial attacks, model vulnerabilities, and implementation examples.
attacks_against_ML_Models
Semester Project
Goal
Implement membership inference attacks against the given pre-trained target models.
🕵️♂️ Membership Inference Attacks on Pre-Trained Neural Networks
This project implements Membership Inference Attacks (MIAs) on pre-trained neural networks using the PyTorch framework. It evaluates privacy leakage vulnerabilities in ML models trained on CIFAR10 and Tiny ImageNet datasets using ResNet34 and MobileNetV2 architectures.
📌 Goal
To assess whether a given data point was part of a model's training set (member) or not (non-member), thereby evaluating potential privacy risks of deployed machine learning models.
🛠 Setup
- Python 3.8+
- PyTorch
- NumPy
- Pickle
- TorchVision
Clone the repo and set up the environment:
bash git clone https://github.com/yourusername/membership-inference-attack.git cd membership-inference-attack pip install -r requirements.txt
Supported Datasets
CIFAR10
Tiny ImageNet
Data Files per Task
shadow.p – Used to train the shadow and attack models.
eval.p – For evaluating attack model performance.
test.p – Final test file for prediction and submission.
Each .p file is a pickled list containing:
shadow.p, test.p: [image, label]
eval.p: [image, label, membership] (1=member, 0=non-member)
Target Models
ResNet34
MobileNetV2