SBIR-STTR Award

The large volume of Automatic Dependent Surveillance-Broadcast (ADS-B) data that is produced can overwhelm analysts, motivating development of automated processing. In recent years, Deep neural networks (DNNs) have produced outstanding results in the image processing domain and are thus attractive candidates for automation of ABS-B processing. The proposed WiseOwl aircraft modeling and behavior analysis tool will leverage artificial intelligence and machine learning to discover behavior patterns and detect anomalies of using ADS-B data. The proposed investigation will draw upon the latest work in DNN-based learning, using a hybrid autoencoder and long short-term memory approach to detect anomalous behavior as well as performing a flavor of specific emitter identification to discover potential message spoofing. The WiseOwl system will augment human analysis for enhanced real-time situational awareness and intelligence production.

North Point Defense Inc. (NPD), intends to develop a system, WiseOwl, for processing ADS-B data using Artificial Intelligence/Machine Learning (AI/ML) and Digital Signal Processing (DSP) to discover insights into the data being transmitted and apparent aircraft behavior. ML technologies are highly effective at processing large volumes of data with high accuracy and can automate and/or augment processes which currently require a human operator. NPD will develop innovative algorithms to analyze and discover patterns in a stream of ADS-B messages as well as provide advanced validation of the reported flight data using aircraft reference data and other raw data sources to add fidelity to the automated behavior analysis. Specifically, NPD plans to employ Deep Neural Network (DNN) approaches which are highly accurate and robust when applied to very large datasets. NPD will develop automated rules-based approaches for analyzing expected aircraft kinematics, clustering algorithms to analyze flight behavior, Convolutional Neural Network (CNN) deep learning approaches to spoofed message detection, and correlation of ADS-B tracks with other sensor’s track data. Phase II prototype development will focus on extending the main Phase I research objectives: data cleaning, analysis of self-reported flight behavior characteristics, and detection of spoofed message attack in RF. Additionally, Phase II will augment the WiseOwl system with new capability to gain further insights into aircraft intent via ADS-B data by performing multilateration which will leverage the distributed sensor nature of Cooperative Engagement Capability (CEC), and by using ADS-B data combined with other sources of data for track validation. Phase II development will result in modular, easy to integrate components that are computationally efficient, accurate, and capable of standing alone or deployed as a complete WiseOwl system. The Phase II prototype system is expected to be completed in 18 months with an exit TRL The WiseOwl system will concisely present discoveries, anomalies, and insights from the ADS-B data. For the prototype system, an easy to use web-based visualization tool will be leveraged; however, in a deployed environment such as CEC, NPD engineers will present data and discoveries in whatever format is necessary for the transition platform. WiseOwl will augment human analysis for enhanced real-time situational awareness and intelligence production. The ability to discover and view flight corridors, abnormal flight behavior, and anomalous ADS-B message traffic will equip commanders with the necessary tools to help assess friend or foe and assist in determining aircraft intent in nearby airspace in tools such as Composite ID, part of the CEC platform.