Date of Award

August 2024

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

Prakash Ranganathan

Abstract

Automatic Dependent Surveillance-Broadcast (ADS-B) is part of the Federal Avi-ation Authority’s (FAA) NextGen position, navigation, and timing (PNT) system, but its packets are unauthenticated and vulnerable to attack by bad actors. This re- search aims to develop methods to detect and mitigate GPS and ADS-B risks for aircraft systems. First, a Random Forest Classifier (RFC) model was trained to clas- sify categories (erroneous, noise, dropouts) of data, achieving an accuracy of 87%. Following this, many single-stage and two-stage forecasting models were tested to forecast Navigation Integrity Category (NIC) values, with the two-stage Random For- est Regressor (RFR) achieving the lowest error or residuals. Next, an interference event lasting 36 hours near the Dallas-Fort Worth (DFW) airport in Texas, USA was identified using an exploratory analysis of both national aircraft data and filtered data within 40 nautical miles (NM). Many possible causes of the interference event were ruled out, with the most likely explanation being human intervention. A novel ensemble S-Value (Signal Value) metric was created to measure inter- ference across multiple features, and then further refined by conducting a feature importance study on a RFR model using SHapely Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) values as the mea- surements of feature importance. The results show that the metrics Navigation Ac- curacy for Velocity (NAC V), Received Signal Strength Indicator (RSSI), Navigation Integrity Category for Barometric Altitude (NIC BARO), Navigation Accuracy for Po- sition (NAC P), and NIC are the most reliable predictors of S-Value.

Download

Share

COinS