TL;DR
This paper tackles the challenge of detecting malicious aircraft location spoofing attacks on ADS-B systems. It proposes a novel method that cross-checks estimated and received ADS-B trajectories using crowdsourced receiver data, even without precise receiver locations or synchronized timestamps, offering a more accessible approach to securing air traffic.
What Problem Does This Paper Address?
The Automatic Dependent Surveillance-Broadcast (ADS-B) system is crucial for air traffic control, providing aircraft with their location and other vital information. However, ADS-B is inherently vulnerable to spoofing attacks, where an aircraft's reported location is maliciously altered. These attacks could potentially lead to dangerous situations, including collisions. While countermeasures have been explored for over a decade, many existing solutions rely on precise receiver locations or synchronized timing, which are often unavailable in crowdsourced networks like the OpenSky Network. This paper addresses the need for a robust and practical method to detect ADS-B location spoofing that can function effectively in such environments.
Methods
The core of the proposed method involves cross-checking the trajectory derived from received ADS-B signals with an estimated trajectory. A key innovation is its ability to work without requiring the exact locations of the ground receivers or precise time-of-arrival information. Instead, it leverages the concept of "sensor coverage" to perform a plausibility check on reported positions. The approach uses a k-NN regression model trained on a grid to estimate aircraft locations. It also incorporates signal strength (RSS) and time-of-flight (ToA) to assess the likelihood of a reported position. The method aims to detect significant deviations in an aircraft's reported trajectory that are unlikely to occur naturally.
Datasets
The paper utilizes data extracted from the OpenSky Network, processing approximately 3 million ADS-B messages. The specific details regarding the dataset's time frame, exact geographical coverage, and how sensor locations were determined or handled are elaborated upon in the responses to reviewer comments, indicating a focus on real-world, unsynchronized data.
Key Insights and Ideas
- Decentralized Security: The method's independence from precise receiver locations and synchronized timestamps makes it particularly well-suited for crowdsourced ADS-B networks, democratizing access to location spoofing detection.
- Plausibility over Precision: By focusing on detecting significant trajectory deviations rather than achieving hyper-accurate localization, the approach offers a practical trade-off for security in environments where precise data is scarce.
- Leveraging Sensor Coverage: The concept of sensor coverage is used as a novel way to validate reported aircraft positions, offering a new angle in the ongoing effort to secure ADS-B.
Key Results
The authors demonstrate that their approach can achieve a precision of less than 20 km, which they argue is sufficient to detect significant trajectory deviations indicative of spoofing attacks. While the exact quantitative performance metrics are detailed in the paper's sections and tables (as referenced in the responses), the core finding is the feasibility of detecting attacks with substantial positional offsets.
Limitations and Future Work
The reviewers highlighted several areas for improvement. A significant point of discussion was the clarity of the "threat model," which the authors have since detailed. The evaluation's focus was initially more on localization accuracy than direct security, prompting clarification and the addition of more detailed metrics. The paper acknowledges that its precision is lower than methods that utilize precise location and timing information. Future work is suggested to address attack detection more comprehensively, potentially by incorporating data from sensors that don't report seeing an aircraft, which could be a stronger indicator of spoofing. The authors also plan to further elaborate on the use of aircraft signals for trajectory derivation and refine the explanation of their weighted localization method.
Summary updated