DeFi Protocol Risk Analysis: An Approach Based on On-Chain Metrics and Classifier Committees

• Edenilson do Nascimento Silva UFERSA
• Isaac Lima Oliveira Filho UERN
• Amanda Gondim de Oliveira UFERSA

DOI: https://doi.org/10.5753/eniac.2026.31450

Resumo

Protocolos DeFi movimentam bilhões de dólares em ativos digitais e permanecem expostos a riscos como colapso de liquidez, exploração de contratos inteligentes e rug pulls. Este artigo investiga métricas financeiras e categóricas on-chain, agregadas ao nível do protocolo, para classificar o risco de protocolos DeFi por meio de Aprendizagem de Máquina. Utilizando dados da API do DeFiLlama, comparamos classificadores individuais, comitês homogêneos e heterogêneos, um meta-classificador via stacking e K-Means. O melhor resultado, um comitê heterogêneo combinando Árvore de Decisão e k-NN, obteve F1-Score de 0,4482 (teste de Friedman, p < 0,05), com viés algorítmico identificado entre blockchains.

Referências

Alhaidari, A., Kalal, B., Palanisamy, B., and Sural, S. (2025). SolRPDS: A dataset for analyzing rug pulls in Solana decentralized finance. In Proceedings of the Fifteenth ACM Conference on Data and Application Security and Privacy, CODASPY ’25, pages 293–298. ACM.

Behfar, S., Mortier, R., and Crowcroft, J. (2026). Probabilistic sampling-enhanced temporal-spatial GCN: A scalable framework for transaction anomaly detection in Ethereum networks. In Proceedings of the Sixth European Workshop on Machine Learning and Systems, EuroMLSys ’26, pages 263–273. ACM.

Green, A. M., Giannattasio, M. P., Erickson, J. S., Seneviratne, O., and Bennett, K. P. (2024). DeFi survival analysis: Insights into the emerging decentralized financial ecosystem. Distributed Ledger Technologies: Research and Practice, 3(1):1–23.

Kong, Q., Chen, J., Wang, Y., Jiang, Z., and Zheng, Z. (2023). DeFiTainter: Detecting price manipulation vulnerabilities in DeFi protocols. In Proceedings of the 32nd ACM SIGSOFT International Symposium on Software Testing and Analysis, ISSTA 2023, pages 1144–1156, New York, NY, USA. ACM.

Luo, B., Zhang, Z., Wang, Q., Ke, A., Lu, S., and He, B. (2024). AI-powered fraud detection in decentralized finance: A project life cycle perspective. ACM Computing Surveys, 57(4):1–38.

Nguyen, M. H., Huynh, P. D., Dau, S. H., and Li, X. (2023). Rug-pull malicious token detection on blockchain using supervised learning with feature engineering. In Proceedings of the 2023 Australasian Computer Science Week, ACSW ’23, pages 72–81. ACM.

Ren, S., He, L., Tu, T., Wu, D., Liu, J., Ren, K., and Chen, C. (2025). LookAhead: Preventing DeFi attacks via unveiling adversarial contracts. Proceedings of the ACM on Software Engineering, 2(FSE):1847–1869.

Tran, M. T., Killeen, B., and McGrath, T. (2026). Adaptive detection of DeFi SLID scams: A data-driven and industry-oriented framework for large-scale DeFi security. In Proceedings of the 2026 Australasian Information Security Conference, AISC ’26. ACM.

Wen, H., Liu, H., Song, J., Chen, Y., Guo, W., and Feng, Y. (2024). FORAY: Towards effective attack synthesis against deep logical vulnerabilities in DeFi protocols. In Proceedings of the 2024 ACM SIGSAC Conference on Computer and Communications Security, CCS ’24, pages 1144–1156, New York, NY, USA. ACM.

Wu, C., Chen, J., Zhao, Z., He, K., Xu, G., Wu, Y., Wang, H., Li, H., Liu, Y., and Xiang, Y. (2024). TokenScout: Early detection of Ethereum scam tokens via temporal graph learning. In Proceedings of the 2024 ACM SIGSAC Conference on Computer and Communications Security, CCS ’24, pages 956–970, New York, NY, USA. ACM.