Smart contracts are programs run automatically on the blockchain and enable peers to enforce agreements without a third-party guarantee. And these contracts must be verified for potential vulnerabilities because they cannot be altered once they are deployed. This paper aims to present a smart contract detection mechanism by integrating a deep learning graph neural network (GNN) model with static and dynamic analysis tools. By combining the strengths of static analysis, dynamic testing, and graph-based deep learning, this work shows how hybrid approaches can be developed to overcome the challenges of smart contract security. By reducing the number of false positives and negatives, this system provides security auditors and developers with an effective way to detect vulnerabilities in smart contracts.