Table 7 Qualitative comparison between the methods presented in previous works and those presented in this work.

LSTM-autoencoders⁶⁸

Time series in the received signal

Software upgrade

Detecting all types of anomalies in the system, not needing to be labeled

Lack of attention to implementation, inability to distinguish the type of attack

BiLSTM-A³⁰

Features available in the Ublox-M8T receiver

Software upgrade

High-accuracy in detection

Not paying attention to the real-time nature of the problem, need for high pre-processing

Federated learning⁶⁹

Spectrogram image of GNSS signal

Software upgrade

Privacy and security, variety of detection

Communication delay, complexity in the model

ResNet⁷⁰

Spectrogram image of GNSS signal

Software upgrade

Adaptability to new scenarios, comprehensive analysis

Computational complexity, extensive pre-processing, the need for dedicated receiver hardware, and implementation challenges

MobileNet-V2²⁷

Spectrogram image of GNSS signal

Software upgrade, additional hardware

Comprehensive coverage of signal disturbance, scalogram-based feature extraction, transfer learning to reduce training time

Heavy computational requirements, dependence on data pre-processing, focus on power without considering environmental factors

RF⁷¹

Features available in the Ublox-M8 receiver

GNSS receiver (M8T), flight controller, raspberry Pi

Hardware implementation, effective model evaluation, minimal performance impact

Reliance on MAVLink protocol, lack of hardware and memory benchmarks, high- weight and dimensions

This work

Doppler, VACC, NUMSV, SNR, SQM, noise level, AGC, jamming inductor, Clk Acc, fix mode

GNSS receiver (M8T), dual frequency active antenna, ARM processor

Real-time detection, low-cost, high-detection accuracy, high-reliability in detection and early warning

Expensive equipment if you need to add more frames