Anomaly detection based on traffic packets in cloud environment

Because the network traffic data is large and the flow rate is high, the abnormal traffic can provide support for the host vulnerability warning, which makes the real-time detection of abnormal network traffic become a key problem. This paper proposes an online anomaly detection model. Firstly, based on the encrypted traffic features of network traffic header, a threshold is set for the feature quantity and the extraction time of each stream to ensure the timeliness of feature extraction. On this basis, an anomaly classification model is constructed based on the classical LeNet-5. Finally, a simulation experiment was carried out to evaluate the anomaly detection accuracy of the above anomaly detection model. The experimental results showed that a small number of header messages could achieve a high anomaly detection accuracy, which provided certain proof for the feasibility of online anomaly detection, also provided support for the subsequent host vulnerability warning, and protected the security of cloud platform. Reduce the impact of attack traffic on the server.