Research

My research is focusing on the societal aspects of Machine Learning (ML) such as fairness, privacy, and security of ML models. Here, I briefly describe my research projects. For a complete list of publications, please visit my google scholar profile.


1. Fairness in ML

Machine learning models are trained based on the data collected from multiple demographic groups. These models may inherit biases against the minority groups, i.e., groups contributing less to the training process may suffer higher loss in model accuracy. In order to address the fairness issue, there are three approaches,

i. Pre-processing: modifying the training datasets to remove the discrimination;

ii. In-processing: imposing certain fairness criterion or modifying the loss function during training;

iii. Post-processing: altering the output of an existing algorithm.

Related Publications:

  1. X. Zhang, MM. Khalili, C. Tekin, M. Liu, “Group Retention when Using Machine Learning in Sequential Decision Making: the Interplay between User Dynamics and Fairness”,  NuerIPS 2019. [pdf]

  2. MM. Khalili, X. Zhang, M. Abroshan, S. Sojoudi, “Fairness and Privacy Improvement in Selection Problems”, AAAI 2021. [pdf]


2. Privacy in ML and Data Market

Traditional machine learning algorithms are not able to preserve individuals’ privacy, and the data used in the training process can be inferred from the trained machine learning model. In this sense, we are facing two challenges. First, due to the privacy concerns, individuals are not willing to share their data with institutions interested in analyzing them, and a proper incentive mechanism is needed to encourage users to share their data. Secondly, we have to design machine learning algorithms that use individuals’ data but at the same time, they can preserve users’ privacy.

Related Publications:

  1. MM. Khalili, X. Zhang. M. Liu, “Designing Contracts for Trading Private and Heterogeneous Data Using a Biased Differentially Private Algorithm”, IEEE Access, 2021. [pdf] [Short Version Published in NetEcon 2019]

  2. X. Zhang, MM. Khalili, M. Liu, “Recycled ADMM: Improving the Privacy and Accuracy of Distributed Algorithms”. IEEE Transactions on Information Forensics and Security, 2019. [pdf] [Early Version Published in ICML 2018]


3. Security Economics

Under-investment in security is a serious issue in computer networks. Firms do not exert enough effort toward securing themselves, which leads to a poor state of security and frequent massive data breaches. Therefore, we have to design incentive mechanisms to incentivize firms to improve their security investments. We investigated the use of cyber insurance and resourced pooling as possible mechanisms to incentivize security investment.

Related Publications:

  1.  MM. Khalili, X. Zhang. M. Liu, “Resource Pooling for Shared Fate: Incentivizing Effort in Interdependent Security Games through Cross-investments”, IEEE Transactions on Control and Network Systems, 2020. [pdf][Early Version in NetEcon 2019]

  2. MM. Khalili, M. Liu, S. Romanosky, “Embracing and Controlling Risk Dependency in Cyber Insurance Policy Underwriting”,  Journal of Cyber Security, 2019. [pdf][Early Version in WEIS 2018]

  3. MM. Khalili, P. Naghizadeh, M. Liu, “Designing Cyber Insurance Policies: The Role of Pre-Screening and Security Interdependence”, IEEE Transactions on Information Forensics and Security, 2018. [pdf][Early Version in NetEcon 2018]

Print Friendly, PDF & Email