Analyzing Suspicious Behaviour in Automatic Tracking of Aircraft Data Using Machine Learning and Deep Learning

Abstract

Ensuring the security and reliability of the Automatic Dependent Surveillance–Broadcast (ADS-B) system is essential for modern aviation safety. This research enhances its resilience by drawing on recent advancements in machine learning as well as deep learning, resulting in a robust detection mechanism. The system processes aircraft transmission data to identify deviations in ADS-B parameters and classify them into three categories: normal, potential anomaly, and confirmed anomaly. This classification enables early identification of irregular patterns that may indicate potential threats or abnormal operational behaviour, helping to mitigate faults or disruptions before they escalate. By employing advanced detection algorithms, the proposed approach strengthens ADS-B system security, supporting safer and more dependable air traffic operations.

Introduction

• Avionics system security has been a national focus since 2003 due to its critical role in infrastructure and airspace safety.

• Modern programs like NextGen (U.S.) and SESAR (EU) have modernized air traffic control systems for efficiency but introduced cybersecurity vulnerabilities.

• Systems like ADS-B, which broadcasts real-time flight data, are mandatory since 2020, but are susceptible to threats such as man-in-the-middle attacks, GPS jamming, and spoofing—especially for UAVs.

• Traditional security fixes are often costly or impractical for real-time systems.

• Machine learning (ML) and deep learning (DL) provide efficient, adaptive alternatives for detecting anomalies in ADS-B transmissions.

2. Research Objective

• Improve security and reliability of ADS-B using ML and DL for anomaly detection.

• Main goals:

  1. Collect and preprocess ADS-B data.

  2. Detect irregular flight parameters.

  3. Classify data into normal, potential anomaly, and confirmed anomaly.

  4. Enhance air traffic safety through better surveillance systems.

3. Literature Review Insights

• Prior studies identified vulnerabilities in ADS-B but lacked real-time implementation or practical anomaly detection.

• Some works used LSTM models and CNNs for detecting irregular ADS-B patterns or GNSS interference, but had limitations such as:

  • High false alarm rates

  • Focused on specific anomaly types

  • Inconsistent performance across flight paths

4. Proposed System Overview

• Combines traditional ML models (e.g., Logistic Regression, AdaBoost) with deep learning models (e.g., CNNs, MLPs).

• Hybrid approach improves detection of complex and subtle irregularities.

• Supports real-time anomaly classification in flight operations.

5. Dataset Description

• ADS-B flight records with detailed attributes:

  • Identification & Flight Info (e.g., Aircraft_ID, Squawk_Code)

  • Position Data (latitude, longitude)

  • Altitude & Flight Level

  • Speed, Heading, Vertical Motion

  • Emergency Indicators

  • Anomaly_Label (target variable: normal, potential anomaly, confirmed anomaly)

6. System Architecture & Methodology

• Modular, layered design with the following stages:

  1. Data Collection – Ingest raw ADS-B data.

  2. Data Preprocessing – Clean, transform, and normalize data.

  3. Feature Extraction – Select meaningful indicators from raw data.

  4. ML/DL Modeling – Train models (CNN, MLP, Logistic Regression, AdaBoost) for anomaly detection.

  5. Decision Layer – Classify outputs into three categories:

     • Normal

     • Potential Anomaly

     • Confirmed Anomaly

7. Key Features & Contributions

• Integrated pipeline for real-time anomaly detection.

• Enhanced ADS-B security without disrupting operations.

• Supports scalability and model comparison across ML/DL techniques.

• Includes an interactive web-based interface for monitoring and visualizing results.

Conclusion

This project makes use of a comprehensive dataset of aircraft flight data, which is necessary for creating and assessing sophisticated anomaly detection algorithms. It makes use of deep learning and machine learning methods to identify risky or irregular flight behaviours in real time, which could jeopardize aviation safety. Altitude, speed, heading, and emergency codes are among the key features that are analysed to promote proactive air traffic control and improved operational safety. The intricate structure of the dataset makes it easier to train models effectively, increasing predicted accuracy and allowing for automated surveillance. By improving anomaly detection capabilities, our research ultimately contributes to safer skies by improving the dependability of ADS-B systems and enabling intelligent, automated monitoring.

References

[1] M. Strohmeier, V. Lenders, and I. Martinovic, “On the Security of the Automatic Dependent Surveillance–Broadcast Protocol,” IEEE Communications Surveys & Tutorials, vol.?17, no.?2, pp.?1066–1087, 2015. DOI: [2] C. Habler and Y. Shabtai, “Using LSTM Encoder-Decoder Algorithm for Detecting Anomalous ADS-B Messages,” Computers & Security, vol. 78, pp. 155–173, Sept. 2018, DOI: https://doi.org/10.1016/j.cose.2018.07.004 [3] M. Riahi Manesh and N. Kaabouch, “Analysis of Vulnerabilities, Attacks, Countermeasures and Overall Risk of the Automatic Dependent Surveillance-Broadcast (ADS-B) System,” International Journal of Critical Infrastructure Protection, vol. 19, pp. 16–31, Dec. 2017, DOI: https://doi.org/10.1016/j.ijcip.2017.10.002 [4] M. Pirolley, R. Couturier, M. Salomon, and F. Ambert, “[Poster] ADS B anomaly detection in the surveillance of low altitude aircrafts,” Journal of Open Aviation Science, vol.?1, no.?2, 2023. DOI: https://doi.org/10.59490/joas.2023.7200

Copyright

Copyright © 2025 Rizwana Begum, Dr. M. Dhanalakshmi. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.