Cache pollution attack detection method based on federated learning under ultra-dense network

Abstract

The invention belongs to the technical field of information security, and provides a cache pollution attack detection method based on federated learning under an ultra-dense network. Firstly, isolatedsmall base stations which are not adjacent to any cluster in the network calculate a weighted distance and decide whether to form a cluster independently, and small base stations which are adjacent to the cluster calculate distance similarity and load similarity and select appropriate clusters to join. And then, each small base station carries out data statistics according to the received interest packet and sends the data statistics to the cluster head, the cluster head is used as a working node in federated learning to be responsible for integrating data to carry out training of a local classifier, and the macro base station is used as a parameter server to be responsible for aggregating the received local classifier to construct an improved global classifier. Finally, the final globalclassifier is broadcasted to all small base stations, the small base stations classify the content by using the classifier after receiving the interest packet, and the content requested by the malicious interest packet cannot be cached and the popularity cannot be updated.

Description

technical field [0001] The invention relates to a cache pollution attack detection method based on federated learning under an ultra-dense network, belonging to the technical field of information security. Background technique [0002] As one of the key technologies of 5G, ultra-dense network (Ultra-Dense Network) can further shorten the distance between end users and access nodes by deploying dense small base stations in hot spots such as shopping malls or transportation hubs, thereby providing low-latency , high data rate, real-time transmission and other communication services. However, the traditional communication method based on the TCP / IP protocol is mainly designed for the host, which can no longer meet the content needs of a large number of users in the mobile network environment. With the rapid increase of traffic from small base stations to the core network, the wireless backhaul link may It will become the performance bottleneck of the network. In order to solv...

AI Technical Summary

Problems solved by technology

As a result, the statistics collected by a single small base station cannot accurately capture the characteristics of abnormal interest packets

In addition, there are some special situations in life. A large number of users suddenly become interested in content that was not popular in the past. At this time, the requests of these legitimate users should be responded normally, but most of the current research judges these interest packets as malicious interest packets. And directly discarded, which seriously damaged the service experience of legitimate users

Images