Design And Implementation Of IoT DDoS Attack Traffic Detection Algorithm Based On Deep Learning

With the development of information and technology,Internet of Things(IoT)devices are widely-spread.However,the security issues of IoT devices are usually ignored.Hackers are good at using system vulnerabilities to infect IoT devices,then build botnets and conduct distributed denial-of-service(DDoS)attacks.Therefore,the security issues of IoT needs to be solved,and the traffic detection algorithm also needs to be updated.At present,machine learning is rarely used in IoT DDoS attacks traffic detection field.To supplement the research gap,this thesis systematically studies the performance of deep learning in the field of IoT DDoS attacks traffic detection.The specific research work is as follows:(1)The performance of lightweight convolutional neural networks(CNN)in the field of IoT DDoS attacks traffic detection is explored,and on this basis,combined with the traffic characteristics of IoT devices,a lightweight IoT traffic detection model Page-Net is designed.This model can reasonably lay out network parameters according to the distribution characteristics of traffic characteristics,and achieve high detection accuracy with a small number of parameter scales,which is more suitable for deployment in edge environments.(2)The method of automatic feature extraction in the field of traditional Internet DDoS traffic detection is introduced into the IoT field and then improved.Specifically,this thesis has designed three deep learning models,including ResC3D,M-ConvLSTM and S-ConvLSTM,to solve the shortcomings of the traditional method of using Long Short-Term Memory(LSTM)and CNN in feature extraction,and realized the synchronous extraction of time and space features.The ResC3D model has the best detection performance by introducing a residual structure in the 3D-CNN model.The M-ConvLSTM and S-ConvLSTM models have obvious advantages in parameter scale and detection time.Therefore,the above model has the advantages of multi-field applications.(3)In order to enhance the practicality of the above deep learning model and facilitate the use of scientific researchers,a flow detection software system is designed.The software system integrates automatic feature extraction function,deep learning model retraining function,result display function and node information sharing function,The user only needs to input the test data,and the software system can feed back the test result for the user by calling the automatic feature extraction function module and the detection module.At present,there is no authoritative data set in the field of IoT traffic detection.In order to ensure the objective authenticity of the data used in the research,this thesis builds an experimental environment consisting of more than 20 kinds of IoT devices and collects the traffic generated by these devices as the experimental data set.The experimental data used in this thesis are all based on this data set.