基于迁移学习-元学习的ADS-B攻击分类研究  

Research on ADS-B Attack Classification Based on Transfer Learning Meta Learning

作　　者：李明[1] 秦柳 宫献鑫 马明远 LI Ming;QIN Liu;GONG Xian-xin;MA Ming-yuan(Civil Aviation Flight University Of China,Guanghan 618000,China)

机构地区：[1]中国民用航空飞行学院,四川广汉618000

出　　处：《航空计算技术》2024年第5期110-114,共5页Aeronautical Computing Technique

基　　金：民航局民航安全能力建设项目资助(ASSA2022/14);中国民航飞行学院科研创新团队项目资助(24CAFUC09018)。

摘　　要：准确分类通航领域ADS-B攻击类型,采取预防攻击措施,对保障通航运行安全性具有重要意义。针对通航领域ADS-B攻击数据样本少,提出一种基于迁移学习-元学习的ADS-B攻击分类模型。该模型将迁移学习与深度卷积自编码器相结合,建立ADS-B攻击特征提取模型,提取数据样本的攻击有效特征表示,并运用元学习策略在特征空间中实现ADS-B攻击准确分类。实例研究表明,基于迁移学习-元学习的攻击分类模型可有效分类小样本ADS-B攻击,且正确率在95%以上。Accurately categorizing the types of ADS-B attacks in the field of navigation and taking preventive measures against the attacks are of great significance to guarantee the security of navigation operation.Aiming at the small number of ADS-B attack data samples in the field of navigation,an ADS-B attack diagnosis and classification model based on migration learning meta learning is proposed.The model combines migration learning with deep convolutional self encoder to build an ADS-B attack feature model,extracts the effective feature representation of the attack from the data samples,and applies a meta learning strategy to achieve accurate classification of ADS-B attacks in the feature space.An example study shows that the attack diagnosis classification model based on migration learning meta learning can effectively classify small sample ADS-B attacks with more than 95%correct rate.

关 键 词：迁移学习 深度卷积自编码器 元学习 ADS-B分类 通航 

分 类 号：TN957[电子电信—信号与信息处理]