30-05-2012, 02:49 PM
Network Coding Based Privacy Preservation against Traffic Analysis in
Multi-hop Wireless Networks
Network Coding Based Privacy.pdf (Size: 391.03 KB / Downloads: 33)
INTRODUCTION
Wireless networks, such as Wi-Fi, have been widely deployed in the access network area due to their
benefits such as convenience, mobility, and low cost. However, they still suffer from their inherent
shortcomings such as limited radio coverage, poor system reliability, as well as lack of security and privacy.
Multi-hop Wireless Networks (MWNs) are regarded as such a promising solution for extending the
radio coverage range of the existing wireless networks. System reliability can be improved through
multi-path packet forwarding, which is feasible in MWNs.*
However, there exist many security and privacy issues in MWNs. Due to the open-air wireless transmission,
MWNs suffer from various kinds of attacks, such as eavesdropping, data modification/injection,
and node compromising; these attacks may breach the security properties of MWNs, including confidentiality,
integrity, and authenticity.
PRELIMINARIES
Network Coding Model
Unlike traditional packet-forwarding systems, network coding allows intermediate nodes to perform
computation on input messages, making output messages be the mixture of the input ones. This elegant
principle implies a plethora of surprising opportunities, such as random coding [2]. As shown in Fig. 2,
whenever there is a transmission opportunity on an outgoing link, an outgoing packet is formed by taking
a random combination of packets in the current buffer.
Homomorphic Encryption Function
Homomorphic Encryption Functions (HEFs) have the property of homomorphism, which means operations
on plaintext can be performed by operating on corresponding ciphertext. For example, suppose
E(⋅) is a HEF. It is easy to compute E(x + y) from E(x) and E( y) without knowing the corresponding
plaintext x and y. To be applicable in the proposed scheme, a HEF E(⋅) needs to satisfy the following
properties:
Privacy Measurement
Currently, privacy measurement consists of two main methods: anonymity set and privacy entropy. In
the anonymity set method, members in an anonymity system is assumed to be uniformly distributed, i.e.,
all members are equally possible to be the target user in probability from the perspective of adversaries.
The privacy entropy method can measure the privacy degree more accurately for those systems where the
members are not uniformly distributed.