|
Supratik Bhattacharyya
|
Sprint
Advanced Technology Laboratories
1 Adrian Court
Burlingame, CA 94010, USA
|
supratik AT sprint DOT com (NEW!!)
(650) 375-4124 phone
(650) 375-4079 fax
|
Last Updated: August 29, 2005
|
|
|
|
- Opportunistic Communication Over Multiple Wireless Networks .
This project aims to develop application support for high-bandwidth but
delay-tolerant applications on mobile devices equipped with multiple
radio interfaces, large storage capacity and powerful processors. The
goal is to enable these devices to make opportunistic use of one or
more available access networks such as wireless WANs, intermittent
wireless LANs and short-distance PANs. A Java-based prototype system
has been developed to support functionality such as automatic
discovery of available networks, intelligent policy-driven use of available
access networks, data transfer over multiple access networks in parallel,
resilience to sudden disconnection via session persistence
and interoperability with legacy servers and applications.
Current work is focused on hardening the Java implementation, supporting applications such as blog uploads and RSS news readers, and investigating policies for access network use based on considerations such as power, pricing and application requirements.
This work is being done in collaboration with
Aaditeshwar Seth and S. Keshav at the University of Waterloo Canada and Jason Lebrun at University of California Davis.
Selected Documents:
- A. Seth, S. Keshav, S. Bhattacharyya. "Opportunistic Data Transfer over Heterogeneous Access Networks". Sprint ATL Research Report RR05-ATL-063000.
- A. Seth, S. Bhattacharyya, S. Keshav. "Application Support for Opportunistic Communication on Multiple Access Networks". Submitted for review.
- Continuous Monitoring for Network Security
The goal of this project is to design and build an "always-on"
packet-level monitoring platform (CMON) for high-speed IP backbone
links, and then use the packet stream for detecting anomalies,
unusual events and malicious activities. The monitoring platform has
been designed, implemented and deployed at a few points in Sprint's
IP backbone. Novel algorithms have been developed for profiling
traffic based on applying data-mining and information- theoretic
concepts to packet-headers. An algorithm has also been developed for
detecting worm scanning activities without requiring knowledge about
TCP connections. On-going work focuses on data streaming algorithms
to process and summarize packet payload characteristics at line
speeds, and on improving resource management, configurability and
adaptability of the monitoring platform itself.
I have several collaborators for this project. The header-based
traffic profiling work is in collaboration with Prof. Zhi-Li Zhang and
his student Kuai Xu at the University of Minnesota Twin Cities. The
data streaming algorithms for payload analysis are being developed in
collaboration with Professor S. Muthukhrishnan and his student Andre
Madeira at Rutgers University.
Selected Documents:
- K. To, T. Ye, S. Bhattacharyya. "CMON: A general-purpose
continuous IP backbone traffic analysis platform" Sprint ATL research
report RR04-ATL-110309.
- K. Xu, Z. Zhang, S. Bhattacharyya. "Profiling Backbone Traffic:
Behavior Models and Applications" ACM SIGCOMM 2005. [ PDF]
- K. Xu, Z. Zhang, S. Bhattacharyya. "Reducing Unwanted Traffic in
a Backbone Network" USENIX 2005 Workshop on Steps to Reducing Unwanted Traffic
in the Internet (SRUTI) [PDF]
- K. Xu, S. Bhattacharyya, Z. Zhang. "A Real-Time Network Traffic Profiling System" Submitted for review.
|
|
|
|
|
|