A Human-in-the-Loop Framework for Detecting Logical Vulnerabilities in Smart Contracts

Abstract Blockchain systems face a wide range of security challenges that span from the consensus layer to the application layer. While many studies have examined consensus protocols or smart contract vulnerabilities separately, fewer works have analyzed these issues together or considered the role of human involvement in the detection process. This paper presents a focused study of security challenges in blockchain, progressing from consensus mechanisms to logical vulnerabilities in smart contracts, and examining the potential of Human-in-the-Loop (HITL) approaches. We review the main families of consensus protocols and highlight their characteristic security trade-offs. We then analyze logical vulnerabilities, which remain difficult for purely automated tools to detect reliably due to their context-dependent nature. Next, we propose a simple HITL framework that combines automated analysis with targeted human review. Experiments on the SmartBugs Curated dataset show that the proposed approach improves precision from 0.7125 to 0.8382 and F1-score from 0.8321 to 0.9120, while reducing false positives by 52.2% without any loss in recall. The results suggest that integrating human expertise can meaningfully improve the detection of logical vulnerabilities. The paper also identifies several research gaps and outlines directions for future work.