The extensive application of smart contract technology in the blockchain domain has positioned it as a key component of the digital economy.However, as the application scope of smart contracts expands, security issues have become increasingly evident, resulting in substantial economic losses.Current vulnerability detection methods largely depend on analyzing complete smart contract source code, which struggles to filter out extensive redundant information when identifying individual vulnerabilities.This leads to an insufficient exploration of the characteristics of smart contract code, which in turn results in lower overall accuracy of detection methods.In this context, the study introduces a method that transforms smart contract source code into a control flow graph, filtering out the main basic blocks based on key information about timestamp vulnerabilities.Subsequently, individual operation audits are conducted on these basic blocks, combined with the graphical structure of the control flow graph, utilizing gated graph neural networks for detecting timestamp vulnerabilities in smart contracts.Comparative experiments were performed on two public datasets, and the results indicate that compared to stateof-the-art detection methods, the proposed approach has improved the accuracy and F1 score for detecting timestamp vulnerabilities in smart contracts by 1.65% and 6.51%, providing valuable application prospects for the security of smart contracts.

As blockchain technology progresses, the deployment of smart contracts has become increasingly prevalent.Despite their growing popularity, smart contracts introduce substantial security risks and vulnerabilities.Present detection methodologies predominantly analyze the entirety of a contract's source code, a process that tends to indiscriminately extract uniform features, leading to inadequate detection capabilities.Addressing these concerns, we propose a novel vulnerability detection method for smart contracts, predicated on a mixed channel attention mechanism.This approach focuses on extracting multifaceted features from pivotal code segments of smart contracts, utilizing a combination of spatial and frequency domain channel attention mechanisms.These mechanisms dynamically modulate the weights of feature channels during visualization, allowing the model to allocate greater emphasis to features of higher significance and enhance feature differentiation.This strategy is instrumental in capturing salient information within the data, thereby augmenting the model's performance and accuracy.Empirical results substantiate that our proposed method substantially surpasses the baseline model, evidencing improvements of 5.55%, 5.42%, and 5.49% in the detection of timestamp vulnerabilities, and 4.12%, 4.29%, and 4.20% in the identification of reentrancy vulnerabilities, respectively.Our method offers a significant advancement in the efficacy of smart contract vulnerability detection and presents valuable implications for future research.