Detection of GNSS Interference Using Reflected Signal Ob-servations from the LEO Satellite Constellation

This study proposes and validates a maximum-based Delay-Doppler Map noise floor strategy with a two-tier verification framework for detecting GNSS interference using CYGNSS LEO satellite data, demonstrating superior sensitivity and reliability compared to existing methods in both controlled jamming tests and persistent interference regions.

The Gist

Here is an explanation of the paper using simple language and creative analogies.

🌍 The Big Picture: Listening to the "Echoes" of GPS

Imagine the GPS satellites in the sky are like giant lighthouses beaming light (radio signals) down to Earth. Usually, we look at the light that comes straight to us. But there's a special fleet of satellites called CYGNSS (pronounced "signs") that acts like a periscope. Instead of just looking at the direct light, they look at the reflections bouncing off the Earth's surface (like the ocean or land) to measure things like wind speed.

However, there's a problem. Sometimes, bad actors on the ground turn on radio jammers (like a giant, angry speaker blasting noise) to drown out the GPS signals. This is called Radio Frequency Interference (RFI). If we can't detect this noise, our GPS navigation, planes, and ships could get lost.

🕵️‍♂️ The Problem: The "Diluted Soup" Effect

For a long time, scientists tried to detect this jamming noise by taking the average of the signals they received.

The Analogy:
Imagine you are at a dinner party with four friends (the four signals the satellite receives at once).

• Three friends are eating quiet, normal food (normal GPS signals).
• One friend suddenly starts screaming because they are being attacked by a fly (the jamming signal).

If you take the average of the noise level at the table, the screaming friend gets "diluted" by the three quiet friends. The average noise level goes up just a tiny bit, maybe not enough to trigger an alarm. The old methods (using averages) often missed these attacks because the noise got lost in the mix.

💡 The Solution: The "Maximum" Detective

The authors of this paper proposed a new strategy: Don't look at the average; look at the loudest voice.

The New Strategy:
Instead of averaging the four friends, the new method asks: "Who is the loudest person in the room right now?"

• If any one of the four signals is screaming (high noise), the system flags it immediately.
• This is like a security guard who doesn't care if 99% of the room is quiet; if one person is screaming, they investigate immediately.

This allows them to catch weak or partial jamming that the old "average" method would have missed.

🛡️ The Safety Net: Avoiding False Alarms

But wait! If we just listen for the loudest voice, we might get scared by a loud sneeze or a dropped plate (false alarms caused by random glitches or terrain errors). How do we know it's a real jammer and not just a glitch?

The authors added a two-step "Truth Check":

1. The "Crowd Check" (Multi-Satellite Verification):
   If two or more satellites flying over the same area hear the screaming at the same time, it's definitely a real jammer. Real jammers are loud enough to be heard by multiple satellites. If only one satellite hears it, we need to check further.

2. The "Stamina Test" (10-Second Rule):
   If only one satellite hears the noise, we don't panic immediately. We wait and watch for 10 seconds.

   • The Physics: A satellite flies very fast (about 7 km per second). If a real jammer is on the ground, the satellite will fly over it for a long time (over a minute). The noise should be continuous.
   • The Glitch: If the noise is just a random glitch or a momentary error, it will disappear in a split second.
   • The Rule: If the noise lasts for a full 10 seconds, we confirm it's a real jammer. If it stops after 1 second, we ignore it.

📊 The Results: Catching the Invisible

The team tested this new method in two places:

1. White Sands Missile Range (USA): Where the government was testing GPS jammers.
2. The Middle East: A region known for constant, messy radio interference.

The Results:

• Old Methods (Averages & Kurtosis): Missed many jamming events, especially when the jamming was weak or only affected one of the four signals.
• New Method (Maximum + Safety Checks):
  • Caught 29% more jamming events in the Middle East compared to the old average method.
  • Detected jamming on three specific days at White Sands where the other methods saw nothing.
  • Could spot the jamming starting up before it got fully loud (like hearing the first cough before the person starts screaming).

🚀 Why This Matters

This paper proves that we can use existing satellites (CYGNSS), which were originally built to measure ocean winds, to act as a global security guard for GPS.

By simply changing how we look at the data (listening for the loudest signal instead of the average) and adding a smart "wait-and-see" rule, we can detect dangerous interference much faster and more reliably, without needing to build expensive new ground stations. It's a smarter way to keep our GPS safe from invisible attackers.

Technical Summary

Here is a detailed technical summary of the paper "Detection of GNSS Interference Using Reflected Signal Observations from the LEO Satellite Constellation" by Shin and Son.

1. Problem Statement

Global Navigation Satellite System (GNSS) signals are increasingly vulnerable to Radio Frequency Interference (RFI), such as jamming and spoofing, due to their low transmit power and open signal structure. Existing detection methods face significant limitations:

• Ground-based limitations: Terrestrial backup systems (e.g., eLoran) are costly to maintain and do not actively detect interference but rather bypass it.
• Space-based limitations: Current Low Earth Orbit (LEO) satellite methods, such as NASA's kurtosis-based approach or mean-based noise floor analysis, often suffer from:
  • Signal Dilution: When using the mean of multiple simultaneous reflected signals, localized interference affecting only a subset of signals is averaged out, leading to missed detections.
  • False Alarms/Missed Detections: Kurtosis-based methods are sensitive to elevation errors, while mean-based methods lack sensitivity to weak or partial interference.
  • Lack of Persistence Verification: Transient outliers caused by receiver noise or geometry errors are often indistinguishable from actual RFI.

2. Methodology

The authors propose a novel Maximum-Based DDM Noise Floor Detection Framework utilizing data from the CYGNSS (Cyclone Global Navigation Satellite System) constellation.

Core Detection Strategy

Instead of averaging the Delay-Doppler Map (DDM) noise floor values of the four simultaneous GNSS reflections received by a CYGNSS satellite, the method selects the maximum value among the four.

• Logic: If any of the four simultaneous reflection paths is contaminated by interference, the maximum value will exceed the threshold, whereas the mean might remain below it.
• Threshold: A fixed threshold of 41 dB was established based on statistical analysis of noise floors over the Pacific Ocean (an RFI-free region).

False Alarm Mitigation (Two-Tier Verification)

To prevent false positives from transient noise or receiver artifacts, the method employs a two-step verification process:

1. Multi-Satellite Concurrent Detection: RFI is confirmed if two or more CYGNSS satellites independently detect elevated noise floors in the same geographic region simultaneously.
2. Temporal Persistence Verification: If only one satellite detects a threshold exceedance, the detection is confirmed only if the noise floor remains above the threshold for a continuous 10-second window.
   • Physical Basis: Based on slant-range geometry analysis, a ground-based jammer at 500 km altitude should remain detectable for over 70 seconds as a satellite passes overhead. A 10-second window is a conservative filter that eliminates transient spikes while retaining genuine interference.

Mathematical Formulation

The detection logic is defined by indicator functions checking:

• $L_s(t)$: Satellite location within the region of interest.
• $R_s(t)$: Maximum of four PRN noise floors > 41 dB.
• $Simul(t)$: $\sum R_s(t) \geq 2$ (Multi-satellite check).
• $Persist_s(t)$: $R_s(\tau) = 1$ for all $\tau \in [t-10, t]$ (10-second persistence).
• Final Decision: $RFI_{proposed}(t) = Simul(t) \cup Persist_s(t)$.

3. Key Contributions

• Maximum-Based Approach: First study to utilize the maximum of simultaneous DDM noise floor values rather than the mean, significantly improving sensitivity to partial-channel interference.
• Physically Grounded Persistence Filter: Introduction of a 10-second temporal verification window derived from orbital mechanics (slant-range geometry), providing a rigorous physical basis for distinguishing RFI from transient noise.
• Dual-Environment Validation: The framework was validated against two distinct scenarios:
  1. Controlled Environment: White Sands Missile Range (NOTAM-announced GPS jamming tests).
  2. Persistent RFI Environment: The Middle East (known for chronic, widespread interference).
• Atypical Pattern Detection: Demonstrated the ability to detect "atypical" interference patterns that do not fit standard vertical stripe models, which other methods missed.

4. Experimental Results

The proposed method was compared against NASA's Kurtosis-based method and a conventional Mean-based noise floor method.

Case Study 1: White Sands Missile Range (May 2025)

• Context: GPS jamming tests were conducted on specific dates.
• Performance:
  • The proposed method detected RFI events on May 14, 15, and 20, dates where both the Kurtosis and Mean methods produced negligible detections.
  • DDM analysis confirmed that on these dates, interference was weak or affected only a subset of the four signals, causing the Mean method to fail (dilution effect).
  • The proposed method successfully identified the early onset of gradually intensifying interference.

Case Study 2: Middle East Region (May 1–24, 2025)

• Context: A region with documented persistent RFI.
• Detection Rates (Flagged Epochs):
  • Proposed Method: 62% (212,534 epochs).
  • Mean-Based Method: 46% (156,487 epochs).
  • Kurtosis-Based Method: 33% (113,950 epochs).
• Interpretation: In a persistent interference environment, the proposed method captured a significantly higher fraction of actual interference events (up to 29% improvement over Kurtosis and 16% over Mean) without a corresponding increase in false alarms in RFI-free zones.

5. Significance and Implications

• Global Monitoring Capability: The study establishes a foundation for using CYGNSS (originally designed for ocean wind retrieval) as a global, infrastructure-free RFI monitoring system.
• Enhanced Reliability: By combining maximum-value sensitivity with geometric persistence verification, the method achieves a superior balance between detection sensitivity (finding weak/partial RFI) and reliability (suppressing false alarms).
• Operational Utility: The ability to detect the onset of interference and atypical patterns provides critical lead time for GNSS users and integrity monitoring systems.
• Future Directions: The authors suggest future work on adaptive thresholds (to account for regional noise variations) and cross-platform validation with other GNSS-R missions (e.g., HydroGNSS).

In conclusion, this paper presents a robust, physics-based framework that significantly outperforms existing space-based RFI detection methods, particularly in scenarios involving localized, weak, or gradually intensifying interference.