Multiple Discriminant Analysis/Maximum Likelihood Research Articles

Detecting rogue attacks on commercial wireless Insteon home automation systems

The Internet of Things (IoT) and commercial wireless home automation applications are expanding as technical capability evolves and implementation costs continue to decrease. However, many home automation devices lack robust security and are vulnerable to a multitude of bit-level attacks. This was highlighted during the first successful Insteon network intrusion demonstration that occurred at DEF CON 23 using a Software Defined Radio (SDR) with YARD Stick One devices. In response, Radio Frequency Distinct Native Attribute (RF-DNA) Fingerprinting is introduced here as a counter-hacking approach for augmenting network bit-level Identity (ID) authentication using Physical Layer (PHY) waveform features. An RF-DNA Fingerprinting process is adopted here and applied to wireless Insteon home automation devices. Rogue device detection is addressed using a Multiple Discriminant Analysis/Maximum Likelihood (MDA/ML) ID verification process. Rogue assessments include attacks by like-model Insteon Switch (IS) devices and YARD Stick One SDR devices programmed to present actual (false) bit-level credentials for authorized Insteon devices while functionally controlling the state of an unprotected (no RF-DNA discrimination) targeted end point device. Device classification and Rogue Rejection Rate (RRR) performance is assessed using Time Domain (TD) and Slope-Based Frequency Shift Keyed (SB-FSK) Fingerprinting with features extracted from a variant (data dependent) signal response region. The Rogue Rejection Rate (RRR) for SB-FSK Fingerprinting was superior to TD Fingerprinting and included RRR ≈ 95% for 25 like-model IS attacks and RRR ≈ 100% for 36 YARD Stick One SDR attacks. SB-FSK Fingerprinting is more computationally efficient (70% fewer features) than TD Fingerprinting and provides an added benefit of being implementable using features extracted from variant data dependent FSK signal response regions.

Classification performance using 'RF‐DNA' fingerprinting of ultra‐wideband noise waveforms

Device classification is important in many applications such as industrial quality control, through-wall imaging and network security. A novel approach has been proposed to use a digital noise radar (DNR) to actively interrogate microwave devices and classify defective units using ‘radio frequency distinct native attribute (RF-DNA)’ fingerprinting and various classifier algorithms. RF-DNA has previously demonstrated ‘serial number’ discrimination of numerous passive radio frequency signals, achieving classification accuracies above 80% using multiple discriminant analysis/maximum likelihood (MDA/ML) and generalised relevance learning vector quantisation-improved (GRLVQI) classifiers. It has also demonstrated above 80% classification of limited active interrogation responses with a DNR signal using these classifiers. The performance capabilities of the two different classifiers, MDA/ML and GRLVQI, on RF-DNA fingerprints produced from the ultra-wideband noise radar correlation response is expanded.

“RF-DNA” Fingerprinting for Antenna Classification

Device classification is important for many applications such as industrial quality controls, through-wall imaging, and network security. A novel approach to detection is proposed using a random noise radar (RNR), coupled with Radio Frequency “Distinct Native Attribute (RF-DNA)” fingerprinting processing algorithms to non-destructively interrogate microwave devices. RF-DNA has previously demonstrated “serial number” discrimination of passive Radio Frequency (RF) emissions such as Orthogonal Frequency Division Multiplexed (OFDM) signals, Worldwide Interoperability for Microwave Access (WiMAX) signals and others with classification accuracies above 80% using a Multiple Discriminant Analysis/Maximum Likelihood (MDAML) classifier. This approach proposes to couple the classification successes of the RF-DNA fingerprint processing with a non-destructive active interrogation waveform. An Ultra Wideband (UWB) noise waveform is uniquely suitable as an active interrogation method since it will not cause damage to sensitive microwave components and multiple RNRs can operate simultaneously in close proximity, allowing for significant parallelization of detection systems.

4G Security Using Physical Layer RF-DNA with DE-Optimized LFS Classification

Wireless communication networks remain underattack with ill-intentioned “hackers” routinely gaining unauthorized access through Wireless Access Points(WAPs)–one of the most vulnerable points in an informationtechnology system. The goal here is to demonstrate thefeasibility of using Radio Frequency (RF) air monitoring to augment conventional bit-level security at WAPs. The specific networks of interest are those based on Orthogonal Frequency Division Multiplexing (OFDM), to include 802.11a/g WiFi and 4G 802.16 WiMAX. Proof-of-concept results are presented to demonstrate the effectiveness of a “Learningfrom Signals” (LFS) classifier with Gaussian kernel bandwidth parameters optimally determined through DifferentialEvolution (DE). The resultant DE-optimized LFS classifier is implemented within an RF “Distinct Native Attribute” (RFDNA) fingerprinting process using both Time Domain (TD) and Spectral Domain (SD) input features. The RF-DNA isused for intra-manufacturer (like-model devices from a given manufacturer) discrimination of IEEE compliant 802.11a WiFi devices and 802.16e WiMAX devices. A comparative performance assessment is provided using results from the proposed DE-optimized LFS classifier and a Bayesian-based Multiple Discriminant Analysis/Maximum Likelihood (MDA/ML) classifier as used in previous demonstrations. The assessment is performed using identical TD and SD fingerprint features for both classifiers. Finally, the impact of Gaussian, triangular, and uniform kernel functions on classifier performance is demonstrated. Preliminary resultsof the DE-optimized classifier are very promising, with correct classification improvement of 15% to 40% realized over the range of signal to noise ratios considered.

Radio frequency fingerprinting commercial communication devices to enhance electronic security

There is a current shift toward protecting against unauthorised network access at the open systems interconnection physical layer by exploiting radio frequency characteristics that are difficult to mimic. This work addresses the use of RF 'fingerprints' to uniquely identify emissions from commercial devices. The goal is to exploit inherent signal features using a four step process that includes: 1. feature generation, 2. transient detection, 3. fingerprint extraction and 4. classification. Reliable transient detection is perhaps the most important step and is addressed here using a variance trajectory approach. Following transient detection, two fingerprinting and classification methods are considered, including 1. power spectral density (PSD) fingerprints with spectral correlation and 2. statistical fingerprints with multiple discriminant analysis-maximum likelihood (MDA-ML) classification. Each of these methods is evaluated using the 802.11a orthogonal frequency-division multiplexing (OFDM) signal. For minimal transient detection error, results show that amplitude-based detection is most effective for 802.11a OFDM signals. It is shown that MDA-ML classification provides approximately 8.5-9.0% better classification performance than spectral correlation over a range of analysis signal-to-noise ratios (SNRA) using three hardware devices from two manufacturers. Overall, greater than 80% classification accuracy is achieved for spectral correlation at SNRA > 6 dB and for MDA-ML classification at SNRA > -3 dB.