基于深度可分离卷积的物联网设备识别模型

IoT Device Recognition Model Based on Depthwise Separable Convolution

随着物联网设备数量的不断增长,物联网设备管理问题逐渐突出,如何在资源有限的物联网环境中准确地识别物联网设备是亟需解决的关键问题。针对物联网设备流量特征提取难的问题,文章提出了一种基于深度可分离卷积的物联网设备识别方法。该方法在会话粒度下利用载荷数据构造设备指纹,通过卷积层从设备指纹中提取深度特征。实验结果表明,该方法能在有限资源下有效识别设备类型。与标准CNN方法和人工特征提取技术相比,整体性能有所提高。

With the continuous growth of the number of IoT devices,the problem of IoT device management has become increasingly prominent.How to accurately identify IoT devices in the resource-limited IoT environment is a key problem to be solved urgently.To solve the difficulty in extracting the traffic features of devices in the Internet of Things(IoT),an Internet of Things device identification method based on deep separable convolution was proposed.In this method,device fingerprints were constructed using payload data at session granularity,and depth features were extracted from device fingerprints through convolutional layer.Experimental results show that this method can effectively identify device types with limited resources.Compared with the standard CNN method and manual feature extraction technique,the overall performance is improved.