Ethereum phishing account detection method based on deep learning

Abstract

The invention provides an Ethereum phishing account detection method based on deep learning, and the method combines transaction time and weight features, enables an Ethereum transaction network to be modeled into a time sequence weighted directed graph, and can effectively capture the properties of a more comprehensive dynamic transaction network. Then, a ski-gram model is adopted to obtain graph embedded feature vectors of all account nodes in the time sequence weighted directed graph, and compared with traditional manual features, the graph embedded feature vectors obtained by the model can adaptively capture implicit features among all accounts; and finally, the obtained graph embedding feature vector is input into a classifier, the classifier effectively combines a k-means clustering algorithm in unsupervised learning and a support vector machine algorithm in supervised learning, and a more robust phishing account classification result can be obtained under the condition of a small amount of label data. According to the method provided by the invention, the phishing accounts can be accurately classified from a large number of accounts, and the ecological security of the Ethereum transaction platform is ensured.

Description

technical field [0001] The present invention relates to the security field of the Ethereum transaction network in the block chain, and more specifically, relates to a method for detecting Ethereum phishing accounts based on deep learning. Background technique [0002] As the underlying technology of digital currency, blockchain technology has the characteristics of data privacy protection, decentralization and non-tampering, and has attracted the attention of researchers all over the world. As a second-generation blockchain platform, Ethereum has smart contract functions and provides a decentralized Ethereum virtual machine through its dedicated cryptocurrency Ether to process point-to-point contracts. Due to the open source nature of Ethereum, all vulnerabilities including security vulnerabilities will be visible. Once these vulnerabilities are exploited by cybercriminals, it will cause great security risks. The more powerful the smart contract, the more complex the logic,...

AI Technical Summary

Problems solved by technology

However, when modeling a directed weighted graph, this method combines multiple transactions between a pair of accounts into one edge, ignoring the diversity and dynamics between accounts, and it is difficult to comprehensively analyze transaction behavior; Two small-scale sub-networks of time series type and network sequence type are used for training; the vertical federated learning strategy is adopted, and the server combines the network embeddings of each edge segment phishing detection model through the attention mechanism; the server uses the combined network embeddings to train itself detection model, and deliver the edge detection model

However, this method uses a fully connected neural network as a classifier when training the phishing detection model, and the effect of the classifier can only be guaranteed on the basis of a large amount of training data. Compared with normal accounts, phishing account nodes in the Ethereum transaction network The number of nodes is very small, and the neural network model cannot obtain robust classification results in the case of unbalanced samples

[0005] In summary, the above phishing account detection methods either ignore the diversity and dynamics among accounts, leading to a one-sided transaction behavior analysis, or do not provide a robust classification model

Images