Abstract: To address the problem that the communication messages of Automatic Dependent Surveillance-Broadcast (ADS-B) system lacked encryption and identity authentication mechanisms and were vulnerable to deceptive attacks such as replay and trajectory spoofing, an ADS-B attack detection method based on an improved DeepLabV3+ model was proposed. First, the original tabular ADS-B data were reconstructed into grayscale images through a grayscale mapping approach, enabling the visualization of spatiotemporal features. Then, to overcome the weak feature extraction capability and blurred segmentation boundaries of the DeepLabV3+ model when processing ADS-B image data, a Haar Wavelet Downsampling (HWD) module was used to replace the max-pooling operation in the ResNet101 backbone network, effectively preserving key features such as trajectory anomalies and local perturbations during dimensionality reduction. Subsequently, a parameter-free attention mechanism (SimAM) was introduced after the low-level feature map to enhance the representation of local spatial structures, and a Fusion Convolutional Block Attention Module (F_CBAM) was incorporated after the ASPP module to highlight critical semantic regions through the parallel interaction of channel and spatial attention. Experimental results on the ADS-B dataset showed that, compared with the baseline DeepLabV3+ model, the proposed method achieved a 6.73% improvement in mean Intersection over Union. In complex attack scenarios, the F1 scores for DoS attacks and random noise reached 96.07% and 92.55%, respectively, which were significantly higher than those of mainstream models such as U-Net, SegFormer, and K-Net. The method provided a new approach for accurate detection of ADS-B attacks.

Key words: Automatic Dependent Surveillance–Broadcast(ADS-B), Attack Detection, Semantic Segmentation, Haar Wavelet Downsampling(HWD), Attention Mechanism

摘要： 针对广播式自动相关监视(ADS-B)系统通信报文缺乏加密与身份认证机制，易受到重放、轨迹伪造等欺骗性攻击，传统检测模型难以精确识别异常的问题，提出一种基于改进DeepLabV3+的ADS-B攻击检测方法。首先，通过灰度图构建方法，将原本的ADS-B表格数据重构为灰度图，实现时空特征的图像化表达；然后，针对DeepLabV3+模型对ADS-B图像数据细节提取能力弱、分割图像边界模糊的问题，采用Haar小波下采样模块(HWD)替代ResNet101主干网络中的最大池化操作，在降维过程中有效保留轨迹异常与局部突变等关键特征；在ADS-B低层特征图后引入无参注意力机制(SimAM)，强化局部空间结构表达，并且在ASPP模块之后引入融合卷积块注意力机制(F_CBAM)，通过通道与空间注意力并行交互突出关键语义区域。实验结果表明，在ADS-B数据集上，与基础DeepLabV3+模型相比，平均交并比提高6.73个百分点；在面对复杂攻击时，尤其是Dos攻击和随机噪声，F1分数分别达到96.07%与92.55%，显著优于U-Net、SegFormer、K-Net等主流模型，为ADS-B攻击的精准检测提供了新思路。

关键词: 广播式自动相关监视, 攻击检测, 语义分割, Haar小波下采样, 注意力机制