Anti-black box detection attack robust electromagnetic signal modulation type identification method

Abstract

The invention discloses a robust electromagnetic signal modulation type identification method against black box detection attack. The method comprises the following steps: constructing a data set through collected electromagnetic signals; establishing a teacher network, training the teacher network through the data set, and generating a soft label through the trained teacher network; and according to a network structure of an actual electromagnetic signal modulation type identification network, building a student network with the same structure, training the student network through the soft label, the data set and the hard label, and transmitting parameters to the electromagnetic signal modulation type identification network. According to the method, the actual electromagnetic signal modulation type identification network parameters are updated, the black box detection attack resisting effect is achieved, the stronger defense capability can be provided when the countermeasure sample is generated for the black box detection attack, the influence of the countermeasure sample on the identification accuracy of the electromagnetic signal modulation type identification network is reduced, and the identification accuracy of the electromagnetic signal modulation type identification network is improved. The deliberate attack of an attacker is overcome, and a more stable electromagnetic signal modulation type identification system is constructed.

Description

technical field [0001] The invention belongs to the technical field based on deep learning and electromagnetic signal modulation type identification, and in particular relates to a robust electromagnetic signal modulation type identification method against black box detection attacks. Background technique [0002] With the continuous development of modern communication systems, the signal environment is becoming more and more complex, and various types of radio equipment are widely used. The types and quantities of radiation sources in the electromagnetic environment are increasing, and the electromagnetic spectrum is becoming more and more crowded. , space, and frequency domains overlap each other. In such a complex scenario, human comprehension can neither complete the perception of a large amount of data, nor can its comprehension speed match the update speed of information, which can easily cause a series of key objectives and situation perception errors or untimely perc...

AI Technical Summary

Problems solved by technology

Meysam Sadeghi and others also verified in the deep learning-based radio signal classification adversarial attack that the adversarial samples are powerfully destructive to the radio signal identification system

Images