Fairness and Security


With the trending discussions on the potential dangers that AI may bring to our society, I carry out research on AI for social goodness. For example, naive automatic decision makings (e.g., for jurying a bail decision, or for making a school admission choice) will possibly reinforce the bias in the training data, which will lead to discriminations. I study fairness in AI systems, and propose algorithms that can guarantee sensible fairness definitions.



Another topic that I found to be extremely important is to protect human subjects' data from being leaked and breached. With technology interacting with our daily life more and more frequently, an increasing amount of our personal data is at the risk of being leaked, e.g., mobile apps gather GPS and use data to track our locations. Potential data (e.g., sensitive information such as credit card information) breach can lead to billions of social losses. In fact, data breach in 2016 leads to a loss of 12 billion US dollars. I developed the first prototype system that leverages on machine learning and data analysis to forecast future data breaches.


Relevant works

Yang Liu, Goran Radanovic, Christos Dimitrakakis, Debmalya Mandal and David Parkes
Calibrated Fairness in Bandits.
FATML 2017, Halifax, Canada.
CODE@MIT 2017, Cambridge, United States.

Yang Liu
Fair Optimal Stopping Policy for Sequential Matching.
UAI 2017, Sydney, Australia.

Yang Liu, Armin Sarabi, Jing Zhang, Parinaz Ardabili, Manish Karir, Michael Bailey and Mingyan Liu
Cloudy with a Chance of Breach: Forecasting Cyber Security Incidents.
USENIX Security 2015, Washington, D.C., United States.

Armin Sarabi, Parinaz Naghizadeh, Yang Liu and Mingyan Liu
Risky Business: Fine-grained Data Breach Prediction Using Business Profiles.
Journal of Cybersecurity, 2016.
Preliminary version appeared at WEIS 2015.

Yang Liu and Mingyan Liu
Detecting Hidden Cliques From Noisy Observations.
IEEE ICASSP 2015, Brisbane, Australia.

On-going works

Bayesian fairness, with Goran Radanovic, Christos Dimitrakakis and David Parkes.

Fairness machine, with David Parkes et al.