Technical Support System

Zibo Wang

318-***-**** *****.***@*****.*** **** Benton St. Apt. B, Ruston, LA 71270

EDUCATION

Ph.D. Degree in Computer Science, Feb 2015 Louisiana Tech University

MS. Degree in Mathematics and Statistics, Feb 2015

MS. Degree in Computer Science, August 2012

B.Eng. Degree in Automation, July 2008 Dalian University of Technology

SKILLS AND STRENGTH

Strong skills and experiences on data mining and machine learning. Great knowledge on algorithms and

schemes in data mining and machines learning. Familiar with related tools, such as SAS, Matlab, R, Excel,

and SQL database. Have work on data as big as hundreds of gigabytes.

Great programming skills:

- Over 10 years of programming experience in C++ and Java.

- Great skill and experience on programming MatLab, SAS, R, SQL.

- Familiar with PHP, Python, Hadoop.

- Great skill in algorithm based programming.

Great work ethic. Can work both individually and within a team.

WORK EXPERIENCE

Louisiana Tech University Ruston, LA

September 2009 – Present

Graduate Research Assistant

Participated in two DARPA-funded projects on active authentication:

1. To improve performance of continuous keystroke verification systems, we designed and developed a “scan-

based” authentication system, to authenticate user identity based on segments of 1 to 5 minutes scans of

continuous biometric data. Using keystroke typing data collected and analyzed from over 400 subjects, we

tested 5 state-of-the-art verification algorithms and fusion techniques in keystroke verification, and evaluated

the “scan-based” authentication system, seeing an improved accuracy over regular mechanisms.

2. We designed and developed a real-time biometric authentication system based on sensors on smartphones,

including touch sensor, accelerometer, and gyroscope. The goal of building such system is to provide a

continuous background user verification on Android smartphones. We combined three biometric modalities in

this system: 1) touch screen swiping (based on touch sensor), 2) touch screen typing (based on touch sensor

and keyboard application), and 3) body movement (based on accelerometer and gyroscope). Our system is

able to verify smart phone user identity continuously based on one or several user modalities. The system was

well tuned to provide better performance, using outlier filtering, feature selection, bootstrapping and so on.

The average error rate of the system was lower than 5%.

Additionally, we designed and developed a robotic attack against smartphones. We wanted to show how a

traditional touch stroke verification system (i.e. based on touch sensor) could be easily broken by a robot.

Two robot models were used: a LEGO EV-3 robot, and a NAO robot. The LEGO EV-3 robot was assembled

with LEGO pieces to build an arm, with a stylus attached as a finger. The finger was programed to move in

three dimensions with three motors. The NAO robot, which was a humanoid robot, could hold stylus on one

hand and a smartphone on the other hand, and was programed to swipe on the touch screen with the stylus.

Both robot was shown to well mimic a human’s swiping pattern, and had 40%-50% rate to break a touch

stroke verification system.

1

The authentication system on smartphones and the robotic attack were demoed in a DARPA meeting, and

received great feedback.

For my Ph.D. thesis, I analyzed user templates evolution in behavior biometric, specifically, in keystroke

dynamics, to evaluate how users’ keystroke behavior changes over time, and how these changes affect the

performance of a keystroke verification system. Results of the analysis indicates that over 40% percent of

keystroke templates changes significantly after a year, and the error rate of a keystroke verification system

increases between 200% and 350%. These results provides solid evidence that template updating mechanism

is needed.

With the previous evidence, I designed and developed an adversary template drifting against template

updating mechanisms. Such adversary attack submits a list of well-tuned biometric samples to stealthily

create fake updates to a victim template, and force the template drifting towards a target template which was

selected by the attacker. The attack was shown to increase system error rate for between 200% and 600%, and

change over 40% of well-performed users to ill-performed users.

To further investigate the dangerous of such attack, I tested the attack in various scenarios to analyze how the

attack perform against defensive mechanism and how defensive mechanism limits the benefit reaped from

template updating.

We build a bench mark test for touch stroke verification on smartphones, to fill gaps of previous testing

mechanisms (i.e. missing outlier filtering, or feature selection). We tested and analyzed 11 state-of-the-art

authentication algorithms on touch stroke verification, with data collected from over 200 subjects, and ranked

these algorithms were ranked based on their error rates.

Lecturer/Teaching Assistant September 2012 - November 2013

Lectured data mining and machine learning to classes of 30-40 senior undergraduate students or graduate

students of computer science program.

Assisted in the teaching and grading for artificial intelligence classes for senior undergraduate students and

graduate students of computer science program.

Technical Support July 2014 - Present

Provided technical support for the college which entailed carrying out system and software upgrades and

maintaining and repairing them as well.

September 2009 – June 2010

Committee, Chinese Students and Scholars Association (CSSA)

Organized activities for Chinese Students and Scholars. Applied and managed funds from the Chinese

Embassy and sponsors. Helped and encouraged new Chinese students for their life in the new country.

PUBLICATIONS

Which Verifiers Work?: A Benchmark Evaluation of Touch-based Authentication Algorithms. In The IEEE Sixth

International Conference on Biometrics: Theory, Applications and Systems (BTAS 2013), 2013.

Scan-Based Evaluation of Continuous Keystroke Authentication Systems. IT Professional, vol.15, no.4, pp.20,23,

July-Aug. 2013

Transforming animals in a cyber-behavioral biometric menagerie with Frog-Boiling attacks. In The IEEE Fifth

International Conference on Biometrics: Theory, Applications and Systems (BTAS 2012), 2012. (One of the four

best reviewed papers)

2

Contact this candidate