Examination of Approaches for Identifying Vulnerabilities in Smart Contracts

Abstract

Objective: By reviewing various previous works, this paper collects the multiple of approaches, strategies used to identify vulnerabilities in smart contracts. Blockchain is a decentralized technology that securely and immutably, records transactions across numerous computers in a visible manner. On a blockchain, smart contracts are self-executing agreements that independently execute and verify contract conditions. This reduces the need for middlemen and increases transparency. Smart contract vulnerabilities are problems in the code that could allow other parties to gain access to, alter, or steal assets as a result of mistakes, faults or imperfections made during development, thereby causing financial and operational harm. In this paper we have algorithms, techniques to detect vulnerabilities in smart contract using deep learning found in literature surveys. Methods: We have found some techniques using opcode, bytecode, Skip-Gram-Word2Vec to convert the smart contract file. Findings: We have found that LSTM, Vanilla-RNN, GRU have very less accuracy 49.64,53.68,54.54. Novelty & Applications: We will come with some different algorithms that will understand different vulnerability with more accuracy. We have come with CNN, Xception, EfficientNet-B2 which has accuracy high then LSTM, Vanilla-RNN, GRU i.e.71,69,75 percent.