Smart contract vulnerability detection method and system, equipment, medium and terminal

Abstract

The invention belongs to the technical field of block chain security, and discloses a smart contract vulnerability detection method and system, equipment, a medium and a terminal. The smart contract vulnerability detection method comprises the steps: taking a large-scale smart contract vulnerability data set as a basic training data set; performing incremental learning training by using the new class of data; dividing all types of test data into a support set and a test set, and inputting the support set data into the trained model to obtain a full-connection network parameter corresponding to each new vulnerability type; taking the full-connection network parameters corresponding to the existing vulnerability type and the new vulnerability type as network nodes, and establishing an observation bypass by using a graph attention network; using a test set to carry out calculation, and predicting calculation cross entropy loss of the result; and performing parameter updating on the graph attention network in the observation bypass by using cross entropy loss, and continuing to train the next batch of new data sets. According to the invention, the continuous learning ability of the model can be trained while the learning vulnerability detection ability is trained based on the old data set.

Description

Technical field [0001] The present invention belongs to the field of block chain security technology, which relates to the field of intelligent contract safety testing, and more particularly to a smart contract vulnerability detection method, system, device, medium, and terminal. Background technique [0002] At present, the intelligent contract is a program that runs on the block chain platform. Like ordinary software, intelligent contracts are also facing various risks. Most of the intelligent contracts on the block chain platform involve digital assets or encrypted currency transactions and processing, due to Use the vulnerabilities of intelligent contracts to attack the block chain platform, which can get huge benefits, resulting in an endless attack on the smart contract vulnerability. In recent years, researchers have begun to use deep learning to safely detect intelligent contracts, and have achieved good results. Currently based on deep learning intelligent contract detec...

AI Technical Summary

Problems solved by technology

[0004] (1) The current smart contract detection model based on deep learning can only be detected based on fixed types of vulnerabilities in the data set, which does not meet the actual application scenarios

[0005] (2) When a newly discovered vulnerability is encountered in an actual application scenario, the current deep learning detection model needs to combine the newly discovered vulnerability data set with the original existing data set for training. This training method is inefficient , in the case of limited computing resources, it is not conducive to the rapid update of deep learning models

[0006] (3) Applying incremental learning to deep learning models will lead to catastrophic forgetting; in the process of incremental learning, only new categories of data sets are used to train deep learning models, and new knowledge will interfere with old knowledge, resulting in Old knowledge is lost, affecting model performance, or even completely losing the ability to classify categories

In addition, incremental learning at this stage requires relatively high experience for researchers

Because before the model is effective, not only a lot of training is required, but also the incremental learning method is very complicated and the training period is very long

Images