Intelligent contract vulnerability detection method based on multi-task learning

The invention discloses an intelligent contract vulnerability detection method based on multi-task learning. An intelligent contract vulnerability detection technology is realized by using a multi-task learning framework based on hard parameter sharing. Firstly, in a data preparation stage, an intelligent contract data set is cleaned, and vulnerability classification and label marking are performed through an existing detection tool; in the data preprocessing stage, an intelligent contract sample source code is compiled to form a byte code, then the byte code is cleaned, then decompiled and converted into an operation code sequence, and input of a model is formed; then, in a model construction stage, constructing an intelligent contract detection model based on multi-task learning; and finally, in a training stage, according to the operation code sequence obtained in the data preprocessing stage, inputting the operation code sequence into the model for training so as to realize judgment and detection of vulne