Biblio

At present, most smart contract vulnerability detection use manually-defined patterns, which is time-consuming and far from satisfactory. To address this issue, researchers attempt to deploy deep learning techniques for automatic vulnerability detection in smart contracts. Nevertheless, current work mostly relies on a single code representation such as AST (Abstract Syntax Tree) or code tokens to learn vulnerability characteristics, which might lead to incompleteness of learned semantics information. In addition, the number of available vulnerability datasets is also insufficient. To address these limitations, first, we construct a dataset covering most typical types of smart contract vulnerabilities, which can accurately indicate the specific row number where a vulnerability may exist. Second, for each single code representation, we propose a novel way called AFS (AST Fuse program Slicing) to fuse code characteristic information. AFS can fuse the structured information of AST with program slicing information and detect vulnerabilities by learning new vulnerability characteristic information.

Automatic vulnerability detection is challenging. In this paper, we report our in-progress work of vulnerability prediction based on graph neural network (GNN). We propose a general GNN-based framework for predicting the vulnerabilities in program functions. We study the different instantiations of the framework in representative program graph representations, initial node encodings, and GNN learning methods. The preliminary experimental results on a representative benchmark indicate that the GNN-based method can improve the accuracy and recall rates of vulnerability prediction.

As everyone knows vulnerability detection is a very difficult and time consuming work, so taking advantage of the unlabeled data sufficiently is needed and helpful. According the above reality, in this paper a method is proposed to predict buffer overflow based on semi-supervised learning. We first employ Antlr to extract AST from C/C++ source files, then according to the 22 buffer overflow attributes taxonomies, a 22-dimension vector is extracted from every function in AST, at last, the vector is leveraged to train a classifier to predict buffer overflow vulnerabilities. The experiment and evaluation indicate our method is correct and efficient.