Classify Encrypted Data in Wireless Sensor Networks
Conference Proceeding Article
End-to-end security mechanisms, like SSL, may seriously limit the capability of in-network processing that is the most critical function in sensor networks. Supporting in-network processing can significantly improve the performance of extremely resource-constrained sensor networks featuring many-to-one traffic patterns. How to protect the traffic and support in-network processing at the same time is an open problem. The paper tackles the problem by proposing a model for categorizing encrypted messages in wireless sensor networks. A classifier, an intermediate sensor node in our setting, is embedded with a set of searching keywords in encrypted format. Upon receiving an encrypted message, it matches the message with the keywords and then processes the message based on certain policies such as forwarding the original message to the next hop, updating and forwarding it or simply dropping it on detecting a duplicate. The messages are encrypted before being sent out and decrypted only at their destinations. Although the intermediate classifiers can categorize the messages, except for several encrypted keywords, they learn nothing about the encrypted messages, not even statistical information. The scheme is efficient, flexible and resource saving. The performance analysis shows that the computational cost and communication cost are minimized. Furthermore, it is resilient to node capture attack and many other kinds of attacks. We are prototyping the model on our mote testbed.
Information Security and Trust
Proceedings of the 2004 IEEE 60th Vehicular Technology Conference, 26-29 September, Los Angeles, California
WU, Yongdong; MA, Di; LI, Tieyan; and DENG, Robert H..
Classify Encrypted Data in Wireless Sensor Networks. (2004). Proceedings of the 2004 IEEE 60th Vehicular Technology Conference, 26-29 September, Los Angeles, California. 3236-3239. Research Collection School Of Information Systems.
Available at: http://ink.library.smu.edu.sg/sis_research/532