MLsploit [Judges Remarks]

Abstract

Machine learning is at risk of being attacked. As companies continue to depend on machine learning to solve their problems, more sophisticated attacks are being created to undermine and take advantage of machine learning algorithms. Worse, these machine learning attacks can have adverse effects on our physical world, like forcing a self-driving car to run a stop sign. MLsploit is a framework designed to solve this problem by allowing operators to evaluate their trained machine learning models against a variety of attacks in order to strengthen them. MLsploit focuses not just on image, video, and audio data, but also contains information security datasets used to detect malware and defend against network intrusions. Using MLsploit, a company can evaluate machine learning models trained on a number of different datasets which provide valuable services to themselves and to their customers.