Meta Neural Trojan Detection

This repo provides an implementation of detecting Trojans in machine learning models as introduced here.

Installation

The code successfully runs on Python 3.6 and PyTorch 1.6.0. The PyTorch package need to be manually installed as shown here for different platforms and CUDA drivers. Other required packages can be installed by:

pip install -r requirements.txt

The MNIST and CIFAR-10 datasets will be downloaded at running time. To run the audio task, one need to download the SpeechCommand v0.02 dataset, extract it under raw_data/speech_command folder and run python audio_preprocess.py. To run the NLP task, one need to download the pretrained GoogleNews word embedding, extract it under raw_data/rt_polarity folder and run python rtNLP_preprocess.py; the movie review data is already in the folder. The Irish Smart Meter Electricity data is private and we do not include them here.

Training Shadow Models and Target Models

The training of shadow models and target models consist of three parts: training the benign models (train_basic_benign.py), training the shadow models with jumbo learning (train_basic_jumbo.py) and training the target models with certain Trojans (train_basic_trojaned.py).

An example of running on the MNIST task:

python train_basic_benign.py --task mnist
python train_basic_jumbo.py --task mnist
python train_basic_trojaned.py --task mnist --troj_type M
python train_basic_trojaned.py --task mnist --troj_type B

Training and Evaluating the Meta-Classifier

run_meta.py trains and evaluates the meta-classifier using jumbo learning and run_meta_oneclass.py trains and evaluates the meta-classifier using one-class learning. An example of training the meta-classifier with jumbo learning on the MNIST task and evaluating on modification attack: