The TrojAI Software Framework: An OpenSource tool for Embedding Trojans into Deep Learning Models
This tool addresses the need for researchers to systematically study and test trojan attacks and defenses in deep learning, though it is incremental as it builds on existing trojan embedding concepts.
The authors introduced the TrojAI software framework, an open-source Python tool for generating poisoned datasets and trojaned deep learning models at scale, and demonstrated its application on MNIST classifiers and a reinforcement-learning model, showing that trigger type, batch size, and poisoning percentage affect trojan embedding, with Neural Cleanse detecting anomalies about 18% of the time.
In this paper, we introduce the TrojAI software framework, an open source set of Python tools capable of generating triggered (poisoned) datasets and associated deep learning (DL) models with trojans at scale. We utilize the developed framework to generate a large set of trojaned MNIST classifiers, as well as demonstrate the capability to produce a trojaned reinforcement-learning model using vector observations. Results on MNIST show that the nature of the trigger, training batch size, and dataset poisoning percentage all affect successful embedding of trojans. We test Neural Cleanse against the trojaned MNIST models and successfully detect anomalies in the trained models approximately $18\%$ of the time. Our experiments and workflow indicate that the TrojAI software framework will enable researchers to easily understand the effects of various configurations of the dataset and training hyperparameters on the generated trojaned deep learning model, and can be used to rapidly and comprehensively test new trojan detection methods.