[Paper Review] An Amateur Drone Surveillance System Based on Cognitive Internet of Things

Guoru Ding, Qihui Wu+4 more|arXiv (Cornell University)|Nov 29, 2017

UAV Applications and Optimization9 references21 citations

TL;DR

This paper proposes Dragnet, a cognitive IoT-based amateur drone surveillance system that enables intelligent detection and classification of authorized versus unauthorized drones using multi-sensor ternary hypothesis testing. By optimizing detection probability under false alarm constraints, the system achieves improved accuracy in identifying unauthorized drones while minimizing false positives, with performance enhancing as sensor count and sampling duration increase.

ABSTRACT

Drones, also known as mini-unmanned aerial vehicles, have attracted increasing attention due to their boundless applications in communications, photography, agriculture, surveillance and numerous public services. However, the deployment of amateur drones poses various safety, security and privacy threats. To cope with these challenges, amateur drone surveillance becomes a very important but largely unexplored topic. In this article, we firstly present a brief survey to show the state-of-the-art studies on amateur drone surveillance. Then, we propose a vision, named Dragnet, by tailoring the recent emerging cognitive internet of things framework for amateur drone surveillance. Next, we discuss the key enabling techniques for Dragnet in details, accompanied with the technical challenges and open issues. Furthermore, we provide an exemplary case study on the detection and classification of authorized and unauthorized amateur drones, where, for example, an important event is being held and only authorized drones are allowed to fly over.

Motivation & Objective

• To address the growing safety, security, and privacy risks posed by unauthorized amateur drones in sensitive airspace.
• To develop a novel, scalable surveillance system tailored for amateur drone monitoring using cognitive IoT principles.
• To enable real-time detection and classification of drones as authorized or unauthorized, particularly in high-security zones like event venues.
• To minimize false alarms while maximizing detection of unauthorized drones through optimized signal processing.

Proposed method

• The system employs a ternary hypothesis testing framework: H₀ (no drone), H₁ (authorized drone), H₂ (unauthorized drone).
• A generalized likelihood ratio test (GLRT) is used to handle the composite hypothesis of unknown unauthorized drone characteristics.
• The optimization problem maximizes Pr(H₂|H₂) subject to constraints on false alarm probabilities Pr(H₁|H₀) + Pr(H₂|H₀) ≤ α and Pr(H₂|H₁) + Pr(H₀|H₁) ≤ β.
• Performance is enhanced by increasing the number of sensors and the number of signal samples, improving global detection probability.
• A genie-aided scheme is used for performance benchmarking, assuming full knowledge of signal distributions from both authorized and unauthorized drones.
• Simulation results evaluate tradeoffs between detection probability and false alarm rates under varying sensor and sample configurations.

Experimental results

Research questions

• RQ1 How can a cognitive IoT framework be adapted to enable effective amateur drone surveillance in high-security environments?
• RQ2 What is the optimal tradeoff between detection probability of unauthorized drones and false alarm rates for authorized and non-existent drones?
• RQ3 How do the number of sensors and sampling duration affect the performance of drone detection and classification?
• RQ4 To what extent can a generalized likelihood ratio test effectively classify unknown unauthorized drones in a composite hypothesis setting?
• RQ5 What are the key technical challenges and open issues in deploying such a system in real-world scenarios?

Key findings

• The detection probability for unauthorized drones (Pr(H₂|H₂)) increases with higher numbers of sensors and longer sampling durations, improving overall system reliability.
• An increase in the number of sensors and samples leads to a significant rise in global detection probability, though marginal gains diminish at high values.
• There exists a fundamental tradeoff between detection performance and false alarm rates, where higher detection requires tolerating increased false positives.
• The proposed generalized likelihood ratio test (GLRT) scheme outperforms baseline methods in detecting unauthorized drones under uncertainty of their signal characteristics.
• The genie-aided scheme, which assumes full knowledge of signal distributions, provides an upper bound on performance, demonstrating the potential gains from perfect channel state information.
• Simulation results confirm that detection performance improves with more samples and sensors, but at the cost of increased detection delay and coordination overhead.