Within the realm of automatic target recognition (ATR) using synthetic aperture radar (SAR), significant research has been performed on the e Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset. The classification results performed on the uncorrupted MSTAR images are typically well above 90% correct, and often approaching 99%. However, in support of operational missions, there is a need to assess the various approaches against a baseline that includes less ideal operating conditions such as foliage penetration (FOPEN). Thus, this paper uses a specialized algorithm that has been proven effective in other settings to assess the effect of a range of increasingly densely spaced pixel amplitude distortions. The results show that once approximately 50% or more of the pixels within the target and shadow region are degraded, the ability to classify the correct target and pose is greatly reduced. Also, as speculated by others, leaving a border of the original clutter appears to yield artificially good classification results in the %50-%90 degraded range before it also rolls off. Finally, when there is no masking, the results are rather sensitive to the chosen confidence level which reinforces the supposition that matches are occurring due to clutter and not just the target.

In support of airborne radar detection missions that rely on Synthetic Aperture Radar (SAR) imagery, there is a need for extensive sets of training data. Due to a paucity of measured data from some targets of interest, there is sometimes a need to train on only simulated SAR data, and yet detect live targets with high confidence during testing. In support of this mission, many researchers have applied a variety of mathematical techniques to simulate data sets. These techniques range from template matching and simpler statistical methods to deep neural networks (DDNs). They demonstrate that with proper pre-processing, some of these methods can achieve target detection with apparently high confidence. However, for all these papers there is no exact measurement of the differences or similarities in the simulated and measured data that would provide a good predictor of the margins between decision boundaries. Thus, this paper has developed a combination of pre-processing methods and standard metrics that enable the assessment of simulated data quality independent of which target recognition algorithm will be utilized. The results show that for some pre-processing methods the differences in simulated data and measured data do not always lend themselves to the desired ability to train on simulated SAR imagery and test on measured SAR imagery.

The Oak Ridge National Laboratory (ORNL) has developed and tested a novel system architecture for acquiring high fidelity high-speed data. The approach uses a consumer grade audio recording device that is normally associated with “garage band” recording of music. ORNL has coupled this low-cost data acquisition hardware with computing technology running open-source software. The main advantage of this approach is per-channel cost; an instrument grade data acquisition system typically costs between $800 to $2000 per channel compared to less than $50 per channel for these consumer grade components. Three systems, each featuring four channels, have been deployed for acquiring data from geophones and the electrical supply system that supports the High Flux Isotope Reactor (HFIR) and the Radiochemical Engineering Development Center (REDC) at ORNL. Each channel samples at 96 kHz at 24-bit resolution. The deployed systems operate continuously 24/7 and produce about 4 terabytes of data per month per system. This paper provides a technical overview of this approach, its implementation, and some preliminary results from qualification testing. This work was conducted in support of the Multi-Informatics for Nuclear Operations Scenarios (MINOS).

This article develops a data -driven, semisupervised approach to learn physical relationships of controller area network (CAN) signals from only a limited set of CAN packets. These mappings are then used to develop a hidden Markov model (HMM) of the driver's actions upon which transaction analysis is performed to optimize the real-time identification of the states. The proposed approach builds an ...

Vehicles are increasingly cyber-physical systems which depend on upwards of 100 or more networked control units. Consequently vehicles, especially those produced after about 2010, face challenges to ensure autonomy, security, and safety. The vehicles’ electronic control units (ECUs) control most of the safety-critical systems. Protecting these networks is especially challenging because there is no publicly available translation of in-vehicle network data to vehicle functions. Thus, an intrusion detection system (IDS) based on mapping the controller area network (CAN) data to 2D images has been developed. While somewhat similar to other recent works that map network features to images, this novel approach utilizes the underlying physical model to automatically group features in a method that makes convolutional neural network (CNN) analysis more feasible. It addresses the most challenging attack in which a compromised ECU sends out incorrect values but sends them within the correct time window. This novel method is shown to detect these kinds of rogue ECU cyber-attacks with greater than a 90% accuracy using very limited training data.

Modern vehicles rely on scores of electronic control units (ECUs) broadcasting messages over a few controller area networks (CANs). Bereft of security features, in-vehicle CANs are exposed to cyber manipulation and multiple researches have proved viable, life-threatening cyber attacks. Complicating the issue, CAN messages lack a common mapping of functions to commands, so packets are observable but not easily decipherable. We present a transformational approach to CAN IDS that exploits the geometric properties of CAN data to inform two novel detectors one based on distance from a learned, lower dimensional manifold and the other on discontinuities of the manifold over time. Proof-of-concept tests are presented by implementing a potential attack approach on a driving vehicle. The initial results suggest that (1) the first detector requires additional refinement but does hold promise; (2) the second detector gives a clear, strong indicator of the attack; and (3) the algorithms keep pace with high-speed CAN messages. As our approach is data-driven it provides a vehicle-agnostic IDS that eliminates the need to reverse engineer CAN messages and can be ported to an after-market plugin.

Vehicle counting, time-of-travel analysis, and other traffic studies frequently require the classification and identification of vehicles in a roadway. Unfortunately, many current technologies for identifying vehicles, such as image-based methods that use cameras and machine vision, are not appropriate for studies that require low-power consumption and low cost. Additionally, privacy issues are becoming a larger concern with the increasing controversy surrounding the public collection of imagery. In this work we evaluate a multi-modal sensing approach to vehicle classification and identification using an ensemble of sensors measurements including electromagnetic emanations and acoustic signatures. A novel kernel regression method is also used for signal learning to classify and identify vehicles without the need of invasive images. Multi-mode sensing, as well as signal learning, is shown to significantly increase the classification rate of specific vehicle classes.

Modern vehicles rely on hundreds of on-board electronic control units (ECUs) communicating over in-vehicle networks. As external interfaces to the car control networks (such as the on-board diagnostic (OBD) port, auxiliary media ports, etc.) become common, and vehicle-to-vehicle / vehicle-to-infrastructure technology is in the near future, the attack surface for vehicles grows, exposing control networks to potentially life-critical attacks. This paper addresses the need for securing the controller area network (CAN) bus by detecting anomalous traffic patterns via unusual refresh rates of certain commands. While previous works have identified signal frequency as an important feature for CAN bus intrusion detection, this paper provides the first such algorithm with experiments using three attacks in five (total) scenarios. Our data-driven anomaly detection algorithm requires only five seconds of training time (on normal data) and achieves true positive / false discovery rates of 0.9998/0.00298, respectively (micro-averaged across the five experimental tests).

Machine learning tools are being developed that support increasingly complex learning-fromsignals on "edge" devices to meet the challenges of decentralized decision making. Edge devices in this context include any electronically enabled device that can sense, process and make decisions based on locally integrated information. Component systems that use algorithms and other technologies are require...

ORNL recently applied its “learning-from-signals” (LFS) techniques to evaluating and improving the energy efficiency of buildings at military installations. LFS is a term coined by ORNL to describe the machine learning algorithms that it has developed for mining, processing, and classifying signals either purposefully or inadvertently being picked up from infrastructure or individual devices. For this particular application, ORNL provided technical support to the Defense Advanced Research Projects Agency (DARPA) Service Chiefs Program for disaggregating electrical power consumption at the device level in a military residential dormitory at Fort Meyer in Washington, DC. The ORNL researchers showed that patterns of device utilization could be monitored on a building's power infrastructure. These devices included cooling/heating water pumps, lighting, washers, dryers, refrigerators, and stoves. This paper discusses the process and initial results of the research effort, as well as the path forward for similar industrial, commercial, and government undertakings.

A limited number of techniques are employed in clinical medicine for regional tissue perfusion assessment. These methods are marginally effective and are not well suited for implantation due to the inability to miniaturize the associated technologies. Consequently, no standardized techniques exist for real-time, continuous monitoring of organ perfusion following transplantation. In this paper, a brief overview of the relevant clinical techniques employed for regional tissue perfusion assessment is given with particular emphasis on post-surgical monitoring of transplanted organs. The ideal characteristics for a perfusion monitoring system are discussed and the development of a new, completely implanted local tissue monitoring system is summarized. In vivo and in vitro data are presented that establish the efficacy of this new technology, which is a photonics-based sensor system uniquely suited for continuous tissue monitoring and real-time data reporting. The suitablity of this sensor technology for miniaturization, which enables implantation for monitoring localized tissue perfusion, is discussed.

The performance and complexity of the signal processing hardware accessible to SDR/CR/RADAR designers has quickly out-paced the available design tools. The advances in Digital Signal Processors (DSP) both fixed- and floating- point, Field Programmable Gate Arrays (FPGA), and multi- core processors have enabled rapid prototyping and deployment of platforms that can be dynamically reconfigured in the field to implement a variety of SDR/CR/RADAR waveforms. Until recently the process of creating waveforms meant starting with high-level mathematical models and simulations and then creating production quality code that can operate on this variety of specialized hardware using either hand coding or vendor specific tools, which are typically limited to single processor solutions. This paper discusses an integrated model-driven design process and tool-flow used in ORNL's Cognitive Radio Program. It describes how the process and tool-flow are used on a variety of SDR and CR projects and in the development of a software-defined RADAR environment simulator. It describes how, from a single Simulink® model, a single deadlock free real-time multi- processor application is created and executed on a network of heterogeneous processors. We also describe recent progress on extending the process/tool-flow to design digital ASICs and our plans for future extensions. We close by highlighting the benefits being realized from applying this design flow to SDR/CR/RADAR projects at ORNL: 1) a significant reduction in the time required to develop, prototype, implement and test SDR/CR/RADAR waveforms, 2) increased reusability/retargetabilty of SDR/CR/RADAR designs and signal processing library components, 3) the ability to quickly port SDR/CR/RADAR waveforms to different hardware systems and processor types, 4) improvements in documentation, and 5) traceability of system components back to original requirements.

The EQ-36 is an enhanced version of the AN/TPQ-36 firefinder radar and is currently under development. In addition to the radar itself being under development, a Radar Environment Simulator (RES) is also under development. The needs for the RES include Testing (including simulated live fire testing), Training, and support of Radar Development. A single solution to answer all of these uses will save time and money and improve soldier proficiency as well as material developer efficiency. Currently an effort is underway with the goal to develop such a multi-use RES. In the past, firefinder RES devices have focused on a subset of these uses and have not attempted to address the broad range of multiple uses targeted by this program. In addition to supporting the EQ-36 RES needs, this RES development effort will strive to maximize reuse of the RES for other firefinder radar models. This paper will provide details on the approach taken as well as status at the time of the symposium.