Table 12 Comparison with existing work.

From: CTDNN-Spoof: compact tiny deep learning architecture for detection and multi-label classification of GPS spoofing attacks in small UAVs

Ref.

Work

Advantages

Result

CTDNN-Spoof

Proposed deep learning method is specifically designed for the detection and multi-label classification of GPS spoofing attacks in Small UAVs, optimizing its effectiveness for this particular application.

Proposed a compact tiny deep learning architecture for detection and multi-label Classification of GPS Spoofing Attacks in Small UAVs

Accuracy: 0.9912, Loss: 0.0027

3

Incorporating explainable artificial intelligence techniques like Shapley Additive Explanations (SHAP), the proposed approach provides insights into why a signal is classified as spoofed. This enhances understanding of the underlying factors contributing to the classification, which can aid in developing more effective mitigation strategies.

Incorporates SHAP to explain why signals are classified as spoofed, providing insights for effective mitigation

F1-score 0.956

1

Self-Supervised Representation Learning (SSRL) integrated with LSTM, GRU, LSTM-RNN, and DNN models to detect GPS spoofing in small UAVs. Incorporates transfer learning to improve adaptability and generalization.

Enhances detection capabilities using SSRL and transfer learning, achieving high accuracy and reduced training time.

Validation Accuracy: 79.0%

Back to article page