Computer Engineering & Computer Science Undergraduate Research Opportunities

Adversarial Machine Learning - Analysis of Black Box Attacks on Machine Learning Tasks

MentorsMehmed Kantardzic and Tegjyot Singh Sethi
Research Lab/LocationData Mining Lab, Duthie Center 241
Description of ResearchWhile machine learning has gained popularity in recent times, its vulnerabilities and securities are just starting to be understood. Adversaries capable of masquerading as end users, can gain access to the system and evade detection by the ML system, using machine learning based systems themselves. We aim to analyze the vulnerabilities of various machine learning tasks, including classification, clustering, recommendation systems and outlier detection systems. This analysis will enable is to better understand the prediction landscape and lead to the development of secure machine learning systems and frameworks.

Analysis of Black Box Attacks
References
  • Sethi, Tegjyot Singh, and Mehmed Kantardzic. "Data Driven Exploratory Attacks on Black Box Classifiers in Adversarial Domains." arXiv preprint arXiv:1703.07909 (2017).
Minimum Student Qualifications
  • Preferred but not required: Experience with Python and Machine Learning techniques.
Pay StatusUnpaid
Timeline & Hours per Week10 hours per week for 8 weeks
Deadline of ProjectNo scheduled deadline
If you are interested