The objective of this research is to develop and analyze new approaches to enhancing security in wireless networks. The approach is to exploit physical layer attributes, in particular the properties of the wireless medium, to significantly enhance user authentication. Openness of wireless and sensor networks makes them vulnerable to spoofing attacks where an unauthorized user masquerades as another legitimate user/device. While conventional cryptographic security mechanisms can be used to foil such attacks, they do not offer a complete solution. This research exploits the distinct channel state information of a legitimate user to authenticate subsequent transmissions from this user.
The intellectual merit of this research includes a thorough consideration of novel efficient approaches to channel estimation for time-varying frequency-selective fading channels under various interference and noise scenarios. The intellectual merit also includes a fairly complete statistical characterization of the estimated channels for use in robust statistical hypothesis testing to ascertain if the estimated physical layer attributes match the previous physical layer attributes from the legitimate user or if they are from a spoofer. Sequential Monte Carlo (bootstrap) approaches in a Bayesian framework are also considered for user authentication.
With respect to broader impacts, this research has the potential to address some of the security challenges that arise from the increasing deployment of wireless networks through enhanced countermeasures. The proposed design methodology for wireless authentication provides a comprehensive framework for enhancing conventional wireless security mechanisms. The project promotes education of students from underrepresented groups and dissemination of research results through teaching at both the undergraduate and graduate levels.
Wireless channel is a challenging communications medium with relatively low capacity per unit bandwidth, random amplitude and phase fluctuations due to multipath time-selective fading, intersymbol interference due to delay spread and multipaths, and interference from other users due to the broadcast nature of the radio channel. The physical link design goal is to achieve data rates close to the fundamental information capacity limits of the channel. Recent results have shown that MIMO (multiple-input multiple-output) channels with multiple transmit and receive antennas are capable of achieving enormous capacity gains over single antenna channels. This has spurred key advances in space-time processing to capitalize on increased Shannon capacity. Accurate knowledge of the CSI (channel state information) of MIMO systems is a prerequisite for most MIMO physical layer approaches. Traditionally a training sequence, in lieu of the information sequence, is transmitted during the acquisition mode to enable the receiver to design an equalizer or estimate the channel in the presence of the aforementioned uncertainties. In the fast time-varying case, the training sequences may have to be transmitted periodically. For a given bandwidth, use of training sequences decreases the effective information rate. In blind channel estimation (system identification) and equalization no training sequences are available or used. In semi-blind channel estimation approaches, a combination of training and information sequence-based data is used so that in addition to the training-based data, one also exploits the information in the rest of the received signal. In superimposed training-based approach the training sequence is ``on'' all the time and is transmitted (at low power) concurrently with (superimposed on) the information sequence. This proposal is concerned with all such three techniques for channel estimation for both single user and multiple users systems and for both time-invariant frequency-selective channels and frequency- and time- selective fading channels.
Identification of fast-varying nonstationary processes is best handled via structured nonstationarities. Our initial focus is on time-varying channels described by a discrete-time complex exponential basis expansion model (CE-BEM) resulting in either a single-input multiple-output (SIMO) time-varying linear system for single user systems or a multiple-input multiple-output (MIMO) linear system for multiuser systems. For wireless channels such canonical models can be derived based on certain physical parameters such as signal bandwidth, channel Doppler spread and multipath spread, up to some unknown time-invariant constants. Other modeling approaches such as wavelet and polynomial bases, will also be considered. We are investigating blind, semi-blind and superimposed training-based system identification techniques for SIMO and MIMO channel estimation, multiuser interference suppression, and equalization and detection of desired user's signal over asynchronous frequency- and/or time-selective fading channels.
Date of latest revision: Tue Aug 5 2008