Qian Xiao

xiaoqian_passport Position Post Doc
Email xiao151@purdue.edu
Office LWSN 2142N

Research Interests & Expertise

My research interests currently focus on data management and analysis, particularly in applying various data models and approximation techniques to practice privacy for real-world data. My work involves privacy-preserving data publishing; randomized approximation algorithms for graph data, high dimensional data and streaming data; network modeling; statistical inference; and access control strategies in online social networks.

Select projects

Anomalous Access Patterns Detection in Relational Databases(2015.8 – Present)

  • Work in a team on using multi-label classifiers to detect anomalous users and anomalous access patterns in relational
    databases
  •  Aim at leveraging the dependencies in relational databases to improve the system’s prediction ability and detect concept
    drift with incremental learning as databases evolve

 

Privacy-preserving High-dimensional Tabular Data Publishing(2014 – 2015)

  •  Worked in a team of three on a light-weight privacy-aware statistical inference framework that uses sampling and junction
    tree algorithm to approximate high dimensional tabular data
  •  Workload: code in R; Github: https://github.com/kaseyxiao/DPTable

 

Privacy-preserving Networking Data Publishing(2011 – 2014)

  • Conducted researches on data anonymization and differentially privacy for privacy-aware networking data publishing
  • Designed two privacy-preserving schemes that utilize the Hierarchical Random Graph model and Markov chain Monte
    Carlo sampling to infer network structure
  • Discovered that approximating network data with connection probabilities can boost data utility while protecting data
    privacy; improved in reducing the magnitude of imposed noise under differential privacy to a logarithmic scale
  • Workload: code in C/C++; Github: https://github.com/kaseyxiao/privHRG

 

Privacy-preserving Streaming Data Publishing(2012 – 2013)

  •  Collaborated with two researchers to design a privacy-preserving streaming data approximation method that assemble
    streaming data with sliding windows on representative streams
  •  Workload: code in Python

 

Access Control in Social Networks(2012)

  • Proposed CAPE, a peer-aware collaborative access control framework that uses game theory and peer effects to facilitate
    social network users in managing privacy settings for group photos
  • Implemented the scheme and deployed a python application as the web server on Heroku; worked with another developer
    to build a Facebook application CAPE-Facebook that connects the web server and Facebook
  • Workload: code in Python with Python Flask framework; Github: http://github.com/kaseyxiao/cape-facebook;
  • Website: http://cape-facebook.herokuapp.com