[Paper Review] A Systematic Literature Review on Smart Contracts Security

Harry Virani, Manthan Kyada|arXiv (Cornell University)|Dec 9, 2022

Blockchain Technology Applications and Security4 citations

TL;DR

This systematic literature review identifies critical security vulnerabilities in Ethereum-based smart contracts, analyzing their root causes and evaluating existing detection techniques. It synthesizes findings from 221 studies to map prevalent flaws—such as reentrancy and integer overflow—and assesses the effectiveness and limitations of static analysis, formal verification, and machine learning tools in mitigating them.

ABSTRACT

Smart contracts are blockchain-based algorithms that execute when specific criteria are satisfied. They are often used to automate the implementation of an agreement so that all parties may be confident of the conclusion right away, without the need for an intermediary or additional delay. They can also automate a process so that the following action is executed when circumstances are satisfied. This study seeks to pinpoint the most significant weaknesses in smart contracts from the viewpoints of their internal workings and software security flaws. These are then addressed using various techniques and tools used across the industry. Additionally, we looked into the limitations of the tools or analytical techniques about the found security flaws in the smart contracts.

Motivation & Objective

• To identify and categorize the most prevalent security vulnerabilities in smart contracts from both internal logic and software implementation perspectives.
• To evaluate the effectiveness and limitations of existing tools and techniques used for detecting smart contract vulnerabilities.
• To provide a comprehensive synthesis of current research on smart contract security to guide future tool development and secure coding practices.
• To highlight gaps in current detection methodologies, particularly in handling complex or novel attack vectors.
• To support developers and researchers in improving smart contract reliability through evidence-based recommendations on tool selection and deployment.

Proposed method

• Conducted a systematic literature review using PRISMA guidelines to identify relevant studies from major academic databases and citation networks.
• Screened 221 studies based on predefined inclusion and exclusion criteria, focusing on empirical research on smart contract security.
• Categorized vulnerabilities into logical flaws (e.g., reentrancy, timestamp dependence) and implementation flaws (e.g., integer overflow, access control issues).
• Mapped detection techniques to vulnerability types, including static analysis, dynamic analysis, formal verification, and machine learning-based approaches.
• Evaluated tools such as Mythril, Slither, and SmartEmbed based on their coverage, precision, and false positive rates.
• Synthesized findings across studies to identify recurring limitations in tooling, such as scalability issues and poor handling of complex control flows.

Experimental results

Research questions

• RQ1 What are the most frequent and critical security vulnerabilities found in Ethereum smart contracts?
• RQ2 Which detection techniques and tools are most effective in identifying these vulnerabilities, and what are their limitations?
• RQ3 How do static analysis, formal verification, and machine learning-based tools compare in terms of accuracy, scalability, and usability?
• RQ4 What gaps exist in current research and tooling for smart contract security, particularly regarding novel or complex attack vectors?
• RQ5 How can developers and researchers leverage existing findings to improve smart contract development practices and tooling?

Key findings

• Reentrancy, integer overflow/underflow, and access control flaws were the most frequently reported vulnerabilities in smart contracts.
• Static analysis tools like Slither and Mythril demonstrated high precision in detecting common vulnerabilities but struggled with complex control flow and indirect calls.
• Formal verification techniques, while highly accurate, were found to be impractical for large-scale contracts due to high resource and expertise requirements.
• Machine learning-based tools such as SmartEmbed achieved high recall in identifying code clones and known vulnerabilities, with a 90% clone ratio observed in 22,000+ Solidity contracts.
• Despite advances, no single tool could comprehensively detect all vulnerability types, highlighting the need for hybrid analysis approaches.
• The study revealed a significant gap in tooling for detecting logic-level attacks such as front-running and oracle manipulation, which remain under-addressed in current research and tools.