Defense method based on signal boundary exploration attack

Abstract

The invention discloses a defense method based on signal boundary exploration attack, which comprises the following steps: S101, screening a to-be-attacked signal A, and carrying out initialization disturbance addition to obtain an adversarial signal A '; S102, according to confidence coefficient changes of the to-be-attacked signal A and the countermeasure signal A ', judging a decision boundarydistribution condition around the to-be-attacked signal A; S103, performing disturbance modification on the countermeasure signal A' according to the decision boundary distribution condition around the to-be-attacked signal A; S104, randomly generating deviation disturbance, so that the adversarial signal A' explores the boundary; S105, optimizing disturbance to enable the countermeasure signal A'not to deviate from an expected track; S106, realizing disturbance optimization until an iteration upper limit is reached or a good confrontation signal A' is obtained; S107, screening adversarial signals, adding the adversarial signals into the training set, and training the signal classification model; and S108, classifying the to-be-classified signals by utilizing the trained signal classification model so as to defend attacks.

Description

technical field [0001] The invention belongs to the research field of deep learning algorithms and data security in the field of artificial intelligence, and specifically relates to a defense method based on signal boundary groping attacks. Background technique [0002] Deep learning can obtain more accurate classification results than general algorithms by learning and calculating the potential connections of large amounts of data, and has powerful feature learning capabilities and feature expression capabilities. The core of deep learning is to use a neural network with huge parameters for feature extraction. Typical neural networks include convolutional neural network (CNN) and recurrent neural network (RNN). [0003] In the past few years, in addition to achieving great success in the fields of computer vision and natural language processing, deep learning technology has also been widely used in other fields, including the use of convolutional neural networks to realize ...

AI Technical Summary

Problems solved by technology

At the same time, compared with some other types of black-box attack methods, one requires less model information and can also be powerful against black-box attacks. Decision-based black-box attacks, such as boundary groping attacks, are easier to attack Machine learning algorithms in the real world interfere, causing errors in signal modulation types and affecting normal life order

Images