Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TRIDENT: Tri-modal Real-time Intrusion Detection Engine for New Targets

Ildi Alla, Selma Yahia, Valeria Loscri

TL;DR

TRIDENT tackles real-time UAV detection under real-world conditions by fusing synchronized audio, visual, and RF data. It introduces two fusion strategies (Late Fusion and GMU) and a synchronized data augmentation pipeline to boost robustness when all modalities degrade simultaneously. The framework demonstrates strong real-world accuracy (up to 96.89% in clean data) and robustness in noisy conditions (83.26% with synchronized test-time noise), while maintaining low latency (about 6 ms per detection) and low energy consumption on edge hardware. A diverse urban/non-urban dataset is released to support reproducibility and future research in robust counter-UAV sensing.

Abstract

The increasing availability of drones and their potential for malicious activities pose significant privacy and security risks, necessitating fast and reliable detection in real-world environments. However, existing drone detection systems often struggle in real-world settings due to environmental noise and sensor limitations. This paper introduces TRIDENT, a tri-modal drone detection framework that integrates synchronized audio, visual, and RF data to enhance robustness and reduce dependence on individual sensors. TRIDENT introduces two fusion strategies - Late Fusion and GMU Fusion - to improve multi-modal integration while maintaining efficiency. The framework incorporates domain-specific feature extraction techniques alongside a specialized data augmentation pipeline that simulates real-world sensor degradation to improve generalization capabilities. A diverse multi-sensor dataset is collected in urban and non-urban environments under varying lighting conditions, ensuring comprehensive evaluation. Experimental results show that TRIDENT achieves 98.8 percent accuracy in real-world recordings and 83.26 percent in a more complex setting (augmented data), outperforming unimodal and dual-modal baselines. Moreover, TRIDENT operates in real-time, detecting drones in just 6.09 ms while consuming only 75.27 mJ per detection, making it highly efficient for resource-constrained devices. The dataset and code have been released to ensure reproducibility (https://github.com/TRIDENT-2025/TRIDENT).

TRIDENT: Tri-modal Real-time Intrusion Detection Engine for New Targets

TL;DR

TRIDENT tackles real-time UAV detection under real-world conditions by fusing synchronized audio, visual, and RF data. It introduces two fusion strategies (Late Fusion and GMU) and a synchronized data augmentation pipeline to boost robustness when all modalities degrade simultaneously. The framework demonstrates strong real-world accuracy (up to 96.89% in clean data) and robustness in noisy conditions (83.26% with synchronized test-time noise), while maintaining low latency (about 6 ms per detection) and low energy consumption on edge hardware. A diverse urban/non-urban dataset is released to support reproducibility and future research in robust counter-UAV sensing.

Abstract

The increasing availability of drones and their potential for malicious activities pose significant privacy and security risks, necessitating fast and reliable detection in real-world environments. However, existing drone detection systems often struggle in real-world settings due to environmental noise and sensor limitations. This paper introduces TRIDENT, a tri-modal drone detection framework that integrates synchronized audio, visual, and RF data to enhance robustness and reduce dependence on individual sensors. TRIDENT introduces two fusion strategies - Late Fusion and GMU Fusion - to improve multi-modal integration while maintaining efficiency. The framework incorporates domain-specific feature extraction techniques alongside a specialized data augmentation pipeline that simulates real-world sensor degradation to improve generalization capabilities. A diverse multi-sensor dataset is collected in urban and non-urban environments under varying lighting conditions, ensuring comprehensive evaluation. Experimental results show that TRIDENT achieves 98.8 percent accuracy in real-world recordings and 83.26 percent in a more complex setting (augmented data), outperforming unimodal and dual-modal baselines. Moreover, TRIDENT operates in real-time, detecting drones in just 6.09 ms while consuming only 75.27 mJ per detection, making it highly efficient for resource-constrained devices. The dataset and code have been released to ensure reproducibility (https://github.com/TRIDENT-2025/TRIDENT).

Paper Structure

This paper contains 33 sections, 10 equations, 16 figures, 17 tables, 2 algorithms.

Table of Contents

  1. Introduction and Motivation
  2. Background on Drone Detection
  3. Threat Model
  4. Dataset Collection
  5. Measurement Locations
  6. Drones and Sensing Equipment
  7. Experimental Conditions
  8. Data Collection Procedure
  9. Sensor Fusion for Drone Detection
  10. Data Pre-processing
  11. Data Augmentation
  12. Unimodal Feature Extraction
  13. Models for Audio Feature Extraction
  14. Models for Visual and RF Feature Extraction
  15. Fusion Strategies
  16. ...and 18 more sections

Figures (16)

  • Figure 1: Measurement locations during data collection.
  • Figure 2: Experimental setup for synchronizing sensor data.
  • Figure 3: Data acquisition method from drones.
  • Figure 4: Audio feature extraction process.
  • Figure 5: T-F representation of drone sound without/with background noise and volume scaling effects.
  • ...and 11 more figures