ADS-B Deceptive Jamming Detection Based On 1D CNN-BiLSTM Network

Among the modern technologies used in air traffic surveillance systems,Automatic Dependent Surveillance-Broadcast(ADS-B)is the most eye-catching one today because its low cost,higher accuracy and Less artificial dependence is widely used by the civil aviation industry.The ADS-B system can accurately extract and process the aircraft’s position,speed,flight number and other identity information,and provide the controller with a clear and intuitive flight path.However,it broadcasts messages without any authentication and encryption,and the attacker can Launch various wireless attacks on the ADS-B system.In recent years,in addition to traditional signal processing methods,deep learning has also been applied to the field of signal processing.Deep learning can automatically extract sample features to complete tasks such as recognition and classification,and shows great advantages,such as modulation signal recognition,signal denoising,radio signal classification,etc.Therefore,this article will introduce deep learning methods to identify ADS-B deceptive interference.The results achieved are as follows:First,because ADS-B signals are periodic and time-related,the problem of detecting ADS-B spoofing signals can be regarded as a timing classification problem.In this paper,the time domain sampling data of ADS-B signal is used as the feature,and two data sets are obtained through data processing.Each training sample of the first data set is an ADS-B signal containing an air position,and the second data set Each training sample is an ADS-B signal sent by the same aircraft within 5 s.Two network models based on 1D CNN-BiLSTM are proposed for two different data sets.In the case of the first data set,a one-dimensional convolutional neural network(1D CNN)preprocesses the received signal first,and then enters Bi-directional long short-term memory network(BiLSTM)to extract the detailed time feature information of the signal;in the case of the second data set,first 1D CNN extracts the detailed time feature information of the ADS-B signal sent by each air position in a flight path,And then use the BiLSTM network to mine the relationship between the ADS-B signals at different air positions.This method does not need to change the existing protocol,nor does it need to decode the message information or calculate the Doppler frequency offset.It only needs to input the original time-domain sampling data as a sample into the network training after preprocessing.Experimental simulation results show that compared with networks that only have time feature extraction,such as DNN,CNN and BiLSTM networks,the two proposed network models have higher recognition accuracy.Second,an improved 1D CNN-BiLSTM network model is proposed to detect ADS-B spoofing interference.Compared with the traditional 1D CNN-BiLSTM network model,the improved model increases the consideration of the relationship between features.After experimental simulation,in the case of two data sets,the improved model has a higher recognition accuracy than the traditional network model.