Design And Implementation Of Security Risk Detection Model On IoT Rules Chain Based On Graph Representation Learning

With the rapid development of the Internet of Things(IoT)and the application of emerging technologies such as 5G and IPV6,more and more items can be connected to each other through the IoT.In the face of these massive IoT devices,a popular solution is using the Trigger-Action Programming(TAP)paradigm to manipulate them.TAP can enable users without programming experience to specify interactive behaviors for IoT devices and services.However,when used by users,the TAP paradigm may inadvertently generate rules chain,that is,rules trigger each other.The security risks caused by the rules chain problem have brought severe challenges to the smart home ecology of IoT.In order to cope with this challenge,this paper proposes an IoT automation home ecological modeling and security risk detection model based on a heterogeneous graph representation learning method.Faced with the uncertainty of rules chain and the security risks they bring,this paper focuses on the textual description in the form of natural language in TAP rules of IoT,and based on this,a reasonable modeling of the rules chain is carried out.This paper uses natural language processing tools and introduces an external knowledge base,extracts nouns and noun phrases in text descriptions as entities,and selects appropriate physical environment factors.On the basis of the construction of IoT rules trigger-action network,IoT rules chain network is constructed,so that the generated heterogeneous network can accurately contain the semantics of rules chain.In order to accurately express the chain relationship between rules,this paper selects the appropriate meta-path in the generated network,then guides the heterogeneous graph random walk algorithm based on the selected meta-path to generate the representation vector of the nodes.Based on the pairwise combination of the node vectors,the representation vectors for rules chain pairs are generated.Based on the existing security level classification method,this paper generates labels with security risk level characteristics for the rules chain pairs in the dataset.This paper conducts multiclassification experiments on rules chain vector pairs,and verifies the accuracy of the heterogeneous graph-based modeling method and meta-path-based graph representation learning algorithm for expressing rules chain semantics,and the feasibility of detecting the security risks of IoT TAP rules chain.This paper conducts multiple experiments on the real IFTTT dataset of IoT to verify the detection model designed in this paper.The experimental results prove the validity and practicability of the detection model proposed in this paper,which is of great significance for the modeling and data processing,security risk analysis and detection of the smart home ecology of IoT in the future.