t i m - w o o d . n e t

 

Sensor Networks

Sensor networks are collections of tiny computers, or "motes", that gather data and transmit it through a wireless network. They are designed to be deployable in hostile environments where human interaction is difficult or impossible. They must use very little power and be able to communicate information even if some nodes in a network fail.



My Project

In 2003-2005 I worked with Dr. Yanyong Zhang, Dr. Wade Trappe, and Wenyuan Xu of the Rutgers WINLAB investigating security issues for wireless networks. Our work focused on detecting and responding to Denial of Service (DoS) attacks on sensor networks.

Wireless networks are prone to many security threats because of their use of a shared media – a small frequency spectrum in the airwaves around us. We have developed methods for detecting jamming style DoS attacks at both the MAC and PHY layers, minimizing the chance that heavy traffic will be mistaken for a malicious adversary.

To defend against DoS, we have designed two strategies to minimize the effects on a network. Following the advice of the military strategist Sun Tzu, we choose to retreat from an enemy that otherwise would overwhelm us. One method, Channel Surfing, is a form of spectral evasion, where network nodes change to a different frequency band beyond the transmission range of the jammer. I have tested this protocol on Berkeley motes, low power sensor nodes with tunable radios, and proven that the devices can detect an attack and effectively switch channels to resume their communication. We intend to develop a wireless ad hoc network that is resilient to attackers, yet uses little overhead so that network throughput can be maintained.

Publications

My first publication was presented at Wise, the Workshop on Wireless Security, at the MobiCom 2004 conference.

A second paper on detection methods for DoS attacks was presented at MobiHoc.

I also contributed to a paper submitted with Jaewon Kang of the Rutgers CS department to Sensys 05 dealing with network congestion in sensor networks.

Motes

The test network I used consisted of MICA2 motes from Crossbow Technologies. Motes are powered by two AA batteries, which give them enough power to last close to a year under regular operation. Each device has a radio transmitter and receiver with a range between 500 and 1000 feet. Programs are writen in the TinyOS language, NesC. Programming is accomplished through a serial connection to a PC running either Linux or the Cygwin emulator.

The complete specifications are available on this data sheet.

t i m - w o o d . n e t