In traditional graph neural neworks, node features are usually transmitted and aggregated through adjacency matrices, but the mask-based smart contract graph node feature enhancement method will cover up some node features during the transmission of the smart contract graph, resulting in loss of available information and causing the subsequent graph convolutional network to over-focus on invalid information during the learning process. In order to better preserve node feature information, this chapter introduces the concept of edge deactivation, which is to make some edge connections in the smart contract graph structure inactive according to the Bernoulli probability distribution during the message transmission process, making the information transmission between nodes more randomized and global. Experimental results show that the smart contract vulnerability detection method based on edge deactivation graph convolutional network has higher detection accuracy than existing methods.