Trevor J. Bihl

Feature Selection for RF Fingerprinting With Multiple Discriminant Analysis and Using ZigBee Device Emissions

The proliferation of low-cost IEEE 802.15.4 ZigBee wireless devices in critical infrastructure applications presents security challenges. Network security commonly relies on bit-level credentials that are easily replicated and exploited by hackers. Unauthorized access can be mitigated by physical layer (PHY) security measures that exploit device-dependent emission characteristics that are sufficiently unique to discriminate devices. RF distinct native attribute (RF-DNA) fingerprinting is a PHY-based security measure, which computes statistical features extracted from such device emissions. However, the RF-DNA fingerprints can be numerous, correlated, and noisy, therefore, a dimensional reduction analysis (DRA) via feature selection is, therefore, of interest. Device classification with DRA feature subsets is evaluated using a multiple discriminant analysis (MDA) classifier. Determining feature relevance from MDA was generally dismissed in prior RF fingerprinting work and is seldom considered in other applications. Here, the MDA feature relevance is revisited using a proposed eigen-based MDA loadings fusion (MLF) methodology. The MDA classification models are adopted and used to assess device identification (ID) classification and verification performance for both the authorized and unauthorized (rogue) devices using a claimed versus actual biometric methodology. Performance is compared for six DRA methods using: 1) a two-sample Kolmogorov-Smirnov test; 2) one-way analysis of variance F-test statistics; 3) a Wilks lambda ratio; 4) generalized relevance learning vector quantized-improved relevance; 5) randomly selected; and 6) the proposed MLF method. Quantitative and qualitative dimensionality assessment methods are compared and contrasted to establish upper bounds on the number of retained features. Experimentally collected ZigBee emissions are considered and ZigBee device classification and ID verification performance using DRA subsets are compared with a full-dimensional feature set. Results show that DRA via the proposed MLF method is superior and more robust than competing methods.

Principal Component Reconstruction Error for Hyperspectral Anomaly Detection

In this letter, a reliable, simple, and intuitive approach for hyperspectral imagery (HSI) anomaly detection (AD) is presented. This method, namely, the global iterative principal component analysis (PCA) reconstruction-error-based anomaly detector (GIPREBAD), examines AD by computing errors (residuals) associated with reconstructing the original image using PCA projections. PCA is a linear transformation and feature extraction process commonly used in HSI and frequently appears in operation prior to any AD task. PCA features represent a projection of the original data into lower-dimensional subspace. An iterative approach is used to mitigate outlier influence on background covariance estimates. GIPREBAD results are provided using receiver-operating-characteristic curves for HSI from the hyperspectral digital imagery collection experiment. Results are compared against the Reed-Xiaoli (RX) algorithm, the linear RX (LRX) algorithm, and the support vector data description (SVDD) algorithm. The results show that the proposed GIPREBAD method performs favorably compared with RX, LRX, and SVDD and is both intuitively and computationally simpler than either RX or SVDD.

Dimensional reduction analysis for Physical Layer device fingerprints with application to ZigBee and Z-Wave devices

Radio Frequency RF Distinct Native Attribute (RF-DNA) Fingerprinting is a PHY-based security method that enhances device identification (ID). ZigBee 802.15.4 security is of interest here given its widespread deployment in Critical Infra-structure (CI) applications. RF-DNA features can be numerous, correlated, and noisy. Feature Dimensional Reduction Analysis (DRA) is considered here with a goal of: (1) selecting appropriate features (feature selection) and (2) selecting the appropriate number of features (dimensionality assessment). Five selection methods are considered based on Generalized Relevance Learning Vector Quantization-Improved (GRLVQI) feature relevance ranking, and p-value and test statistic rankings from both the two-sample Kolmogorov-Smirnov (KS) Test and the one-way Analysis of Variance (ANOVA) F-test. Dimensionality assessment is considered using previous qualitative (subjective) methods and quantitative methods developed herein using data covariance matrices and the KS and F-test p-values. ZigBee discrimination (classification and ID verification) is evaluated under varying signal-to-noise ratio (SNR) conditions for both authorized and unauthorized rogue devices. Test statistic approaches emerge as superior to p-value approaches and offer both higher resolution in selecting features and generally better device discrimination. With appropriate feature selection, using only 16% of the data is shown to achieve better classification performance than when using all of the data. Preliminary first-look results for Z-Wave devices are also presented and shown to be consistent with ZigBee device fingerprinting performance.

Defining, Understanding, and Addressing Big Data

“Big Data” is an emerging term used with business, engineering, and other domains. Although Big Data is a popular term used today, it is not a new concept. However, the means in which data can be collected is more readily available than ever, which makes Big Data more relevant than ever because it can be used to improve decisions and insights within the domains it is used. The term Big Data can be loosely defined as data that is too large for traditional analysis methods and techniques. In this article, varieties of prominent but loose definitions for Big Data are shared. In addition, a comprehensive overview of issues related to Big Data is summarized. For example, this paper examines the forms, locations, methods of analyzing and exploiting Big Data, and current research on Big Data. Big Data also concerns a myriad of tangential issues, from privacy to analysis methods that will also be overviewed. Best practices will further be considered. Additionally, the epistemology of Big Data and its history will be examined, as well as technical and societal problems existing with Big Data.

Towards the mitigation of correlation effects in anomaly detection for hyperspectral imagery

This research involves simulating remote sensing conditions using previously collected hyperspectral imagery (HSI) data. The Reed–Xiaoli (RX) anomaly detector is well-known for its unsupervised ability to detect anomalies in hyperspectral images. However, the RX detector assumes uncorrelated and homogeneous data, both of which are not inherent in HSI data. To address this difficulty, we propose a new method termed linear RX (LRX). Whereas RX places a test pixel at the center of a moving window, LRX employs a line of pixels above and below the test pixel. In this paper, we contrast the performance of LRX, a variant of LRX called iterative linear RX (ILRX), the recently introduced iterative RX (IRX) algorithm, and the support vector data description (SVDD) algorithm, a promising new HSI anomaly detector. Through experimentation, the line of pixels used by ILRX shows an advantage over RX and IRX in that it appears to mitigate the deleterious effects of correlation due to the spatial proximity of the pixels; while the iterative adaptation taken from IRX simultaneously eliminates outliers allowing ILRX an advantage over LRX. Such innovations to the basic RX algorithm allow for the reduction of bias and error in the estimation of the mean vector and covariance matrix, thus accounting for a portion of the spatial correlation inherent in HSI data.

QUEST hierarchy for hyperspectral face recognition

A qualia exploitation of sensor technology (QUEST) motivated architecture using algorithm fusion and adaptive feedback loops for face recognition for hyperspectral imagery (HSI) is presented. QUEST seeks to develop a general purpose computational intelligence system that captures the beneficial engineering aspects of qualia-based solutions. Qualia-based approaches are constructed fromsubjective representations and have the ability to detect, distinguish, and characterize entities in the environment Adaptive feedback loops are implemented that enhance performance by reducing candidate subjects in the gallery and by injecting additional probe images during the matching process. The architecture presented provides a framework for exploring more advanced integration strategies beyond those presented. Algorithmic results and performance improvements are presented as spatial, spectral, and temporal effects are utilized; additionally, a Matlab-based graphical user interface (GUI) is developed to aid processing, track performance, and to display results.

An Optimization Framework for Generalized Relevance Learning Vector Quantization with Application to Z-Wave Device Fingerprinting

Z-Wave is low-power, low-cost Wireless Personal Area Network (WPAN) technology supporting Critical Infrastructure (CI) systems that are interconnected by government-to-internet pathways. Given that Z-wave is a relatively unsecure technology, Radio Frequency Distinct Native Attribute (RF-DNA) Fingerprinting is considered here to augment security by exploiting statistical features from selected signal responses. Related RF-DNA efforts include use of Multiple Discriminant Analysis (MDA) and Generalized Relevance Learning Vector Quantization-Improved (GRLVQI) classifiers, with GRLVQI outperforming MDA using empirically determined parameters. GRLVQI is optimized here for Z-Wave using a full factorial experiment with spreadsheet search and response surface methods. Two optimization measures are developed for assessing Z-Wave discrimination: 1) Relative Accuracy Percentage (RAP) for device classification, and 2) Mean Area Under the Curve (AUCM) for device identity (ID) verification. Primary benefits of the approach include: 1) generalizability to other wireless device technologies, and 2) improvement in GRLVQI device classification and device ID verification performance.

Vibrometry-based vehicle identification framework using nonlinear autoregressive neural networks and decision fusion

This research considers simulated laser radar (LADAR) vibrometry for vehicle identification. Time sampled data is considered for developing multiple nonlinear autoregressive neural network (NARNet) classifier models. Emphasis is placed on robustness to sensor location and using small amounts of data. Decision level fusion is used to combine results from multiple classifiers. Results offer improved classification performance as compared to the literature.

Hybrid System Identification for MIMO Control-System Design

Abstract Herein a method of hybrid frequency domain and state space system identification is developed, tested and implemented. A frequency domain least squares system identification algorithm, along with a coherence function technique for eliminating noisy data is used to sequentially develop discrete single-input, multiple-output (SIMO) transfer function models between each input and the outputs. From the transfer function models, difference equations are obtained. Using the difference equations, discrete impulse responses between each input and each output are computed. These impulse responses are then processed by a state space system identification technique to create a minimum order state space multiple-input, multiple-output (MIMO) model. This process is illustrated on a controls-structures interaction platform called Flexlab.

The effectiveness of using diversity to select multiple classifier systems with varying classification thresholds

In classification applications, the goal of fusion techniques is to exploit complementary approaches and merge the information provided by these methods to provide a solution superior than any single method. Associated with choosing a methodology to fuse pattern recognition algorithms is the choice of algorithm or algorithms to fuse. Historically, classifier ensemble accuracy has been used to select which pattern recognition algorithms are included in a multiple classifier system. More recently, research has focused on creating and evaluating diversity metrics to more effectively select ensemble members. Using a wide range of classification data sets, methodologies, and fusion techniques, current diversity research is extended by expanding classifier domains before employing fusion methodologies. The expansion is made possible with a unique classification score algorithm developed for this purpose. Correlation and linear regression techniques reveal that the relationship between diversity metrics and accuracy is tenuous and optimal ensemble selection should be based on ensemble accuracy. The strengths and weaknesses of popular diversity metrics are examined in the context of the information they provide with respect to changing classification thresholds and accuracies.