Currently, I am a graduate student in Department of Computer
Science, at Iowa State University. My research interests are
wireless networks and mobile computing. What follows is
current and some accomplished work.
Ames, IA:
2004 - Present ------------ Iowa State University
KAA: Location Anonymity in Continuous
Location Based Service Systems. Location based services
(LBS) have become more and more popular with the advanced mobile
devices and wireless communication technology. However, to receive
LBS, the mobile users must submit their location information to
server. Thus, to protect users' privacy, an anonymity area, which
covers K-1 other mobile users, is reported to server instead of
actual position. But current work does not consider the continuous
LBS which lasts a certain period of time. Adversary can compromise
the location privacy by correlating multiple consecutive anonymity
areas. In this project, I introduce the concept of entropy, and give
a new definition of anonymity degree of an anonymity area for
continuous LBS. As a result, the series of anonymity areas generated
by my algorithm are robust under the correlation attack. Related Publicaiton:Toby Xu, Ying Cai, Location Anonymity in Continuous
Location-based Services, ACM GIS 2007, to appear.
GeoNet (ePositit): A Spatial Messaging System in Mobile Wireless Networks.
This project is sponsored by IPRT. It is a wireless
communication system consists of a central server and a number of
mobile users. In this system, the mobile users can mark a geographic
region and post a location-specified message correspondingly. The
message may be an advertisement, a danger warning, a traffic jam
report, etc. When other mobile users enter this region, the message
is transmitted from server to user and shown on the mobile device.
Normally, to test whether the user has entered a message region, it
periodically reports its location during its movement, which may
lead to a communication bottleneck on server. In this project, we provide an efficient technique
alleviate this problem. Each mobile device
(Pocket PC) is assigned a resident domain according to its location
and computation capability. It keeps monitoring its position against
message areas in its resident domain. A mobile device reports to server its new
status only when it crosses the boundary of those message area. As a
response, the sever sends the content of corresponding spatial messages to the
mobile user......(more)
Related Publication: Real-Time Processing of
Range-Monitoring Queries in Heterogeneous Mobile Databases, Ying
Cai, Kien A. Hua, Guohong Cao, Toby Xu. IEEE Transaction on Mobile
Computing. Vol. 5(7).
Page 931~942. 2006 [Abstract]
Technique Report: GeoNet: Where the Internet Meets the Physical
World. Link.
The GUI on mobile user device (HP IPAQ 4310)
Life
streaming in mobile ad hoc networks (MANET): Life
streaming means continuous data delivery. To guarantee a smooth
playback, each data packet must be delivered with a strict deadline.
But in MANET, the network topology keeps changing, such that a
routing path used by the current data packet may become unavailable
for the next one. In this project, we present a light-weighted yet
highly efficient streaming protocol PLP-ROA for MANET. The idea of
PLP is to allow each host in the streaming path to monitor the
robustness of its connection, and proactively looks for a backup
link when the current connection is about to break. The second
technique, ROA, makes it possible to maintain a high-quality
streaming path (having a minimum number of hops) by
dynamically adjusting a streaming path so that it can be as straight
as possible along the direction from the
stream source to the receiver.....(more)
Related publication: Streaming in MANET: Proactive Link
Protection and Receiver-Oriented Adaptation. Toby Xu, Ying Cai, IEEE
IPCCC'07 (Abstract)
Beijing: 2002-2004
-------- Institute of Automation, China Academy of
Science.
National Lab of Pattern Recognition:
Face tracking and recognition:
This project was sponsored by National Natural Science
Foundation of China. I developed a real-time face tracking
and recognition system in office environment.
A camera could capture the people who came into the office,
picture a standard fontal face of the person and
make corresponding identification. During the algorithm design. I
found that the HSV color space is not sensitive for the change of
luminance. This color model is more robust under luminance variance.
On the other hand, Adaboost is a fast for face detection, but it
suffers detection failure when face is not frontal. Thus, for
real-time face detection, I combined the two technique together.
Next, I used EigenFace method for face recognition, in which the
face image were projected onto a face space which consists of a
complete basis of eigenfaces. The images of the same person fell
into a small region in the face space. The experiment got a recognition rate of 95%.
Low
High
The
skin color histogram in H-S space under different luminance
The face detection result of
skin color model under different luminance
Hefei:
2001-2002 -----------
University of Science and Technology of China
National
High Performance Computing Center
Distributed
Scheduling in Linear Networks. Given a group
of job subscribers, each of which generate a number of jobs, it can
distributed the jobs within the networks instead of processing all
the jobs by themselves. The problem is that how to distribute these
jobs in order to minimize the job finishing time. In this project,
we focused on the problem of
distributed job scheduling on a special topology of linear
networks. We first figured out the lower bound
l of the optimal
schedule. Then, we designed a constant rate
r-approximation algorithm
which guarantee that the all jobs can be finished within the time of
r times
l.
Related publication: Distributed Job
Scheduling in Linear Networks, Guangzhong Sun, Guoliang Chen, Chen
Chen, Ge Xu, Zheng Zhang, Journal of Computer Research and
Development, Vol 40. No.10
Oct. 2003,
p1476~p1481. [Abstract]
