Traditional low probability of intercept/low probability of detection (LPI/LPD) radar and communication systems have been fielded since the late 1970s employing both transmit power management and spread spectrum modulation techniques. Early electronic signal measures (ESM) techniques to detect these emissions relied heavily on energy detection primarily due to real-time processing limitations of the technology at that time. Alternatively, spectrum listening and subsequent processing (non real-time) was employed to detect and classify spread spectrum signals termed electronic intelligence (ELINT) to aid in the development of real-time methods to look for specific spread spectrum modulation techniques efficiently. The processing would develop pulse descriptor words and utilize libraries to perform specific emitter identification (SEI). As processing technology advanced the ability to perform ELINT type processing in real-time became increasingly possible using various types of bilinear and linear time-frequency methods producing time-frequency signatures of the spread spectrum modulation that realized partial matched filtering type gain, enabling both detection and extraction of modulation parameters facilitating SEI. The establishment of these capabilities was soon followed by advances in machine learning (ML) and artificial intelligence (AI) techniques that were motivated by image processing for computer vision or object recognition. Hence, time-frequency representations of spread spectrum modulation waveforms constitute an image with distinguishing features suitable for characterization by ML AI methods performing both detection and characterization. More recently, ML AI methods have been scrutinized for their susceptibility to intentional misclassification by adversarial examples, and research is underway to detect adversarial examples. and harden ML AI methods against the misclassification that they can cause (i.e., counter electronic attack. LSI has expertise and experience that span all relevant technology areas of this topic that include detailed knowledge of spread spectrum modulation techniques, time-frequency representations of these modulation techniques, the development and utilization of ML AI techniques that include (neural nets, SVMs, Bayesian methods and deep learning implementations), and the development of adversarial examples, their detection, and hardening ML AI methods against misclassification when presented with adversarial examples. LSI also has detailed knowledge of all radar modalities and the associated propagation and scattering phenomenology and will focus on designing LPI/LPD waveforms for air-to-surface maritime modes for the Phase 1 effort.
Benefit: Promising results could lead to substantial improvements in automatic maritime situational awareness (MSA) and mission prosecution that minimizes detectability and vulnerability to CEA. These capabilities are applicable to a broad class of platforms with diverse sensors and missions made possible by RM/Minotaur (e.g., MQ-4C, P-8A, MH-60R, and MQ-8C).
Keywords: AI, AI, ML, LPD, Radar, LPI
