Leveraging Deep Learning Techniques for Accurate Detection of Critical Vulnerabilities in Solidity Smart Contracts
Open Access DepositedBlockchain is a relatively new technology that has been quickly adopted across many businesses, dramatically increasing the need for organizations to secure their Solidity smart contracts. Smart contracts contain vulnerabilities, and as a result, blockchain applications that support core business processes face significant security and reliability issues. There have been several attempts using conventional methods for detecting vulnerabilities in smart contracts, however, there have been issues with reliability and effectiveness, regardless of ongoing development. This research explores some Deep Learning (DL) techniques and their effectiveness in detecting major vulnerabilities in Solidity smart contracts. There were several models that have been explored in this research, including transformer-based architectures, such as CodeBERT and GraphCodeBERT, along with Bidirectional Long Short-Term Memory (BiLSTM), Convolutional Neural Networks (CNN), and Recurrent Convolutional Neural Networks (RCNN). The research uses a combination of datasets that contain more than 53,000 files of Solidity smart contracts. All these contracts have gone through a solid approach of cleaning and labeling. Additionally, data preprocessing has been applied prior to training each model. This includes Exploratory Data Analysis (EDA), systematic tokenization, as well as feature engineering. This research showed that transformer-based architectures have a promising future as they deliver the best results overall among all metrics, making them achieve remarkable precision and recall scores while keeping performance at a high level for both binary and multi-label classifications. This praxis uses a new DL modeling framework to showcase a scalable and effective modeling approach for security analysis of smart contracts, which has a value-added contribution to the cybersecurity literature and wider security community overall. The paper explored several research hypotheses and answered fundamental research questions using a robust evaluation methodology. Additionally, the research added suggestions to the existing body of knowledge that can be used for future research directions to help enhance the vulnerability detection of smart contracts in real-time scenarios, which can improve blockchain security overall.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
AlMuarrawi_gwu_0075A_17405.pdf | 2025-12-12 | Open Access |
|