Approaches to Security in Distributed Applications

The article examines security approaches in distributed applications (dApps) based on blockchain technologies and related decentralized systems. The primary focus is on existing consensus algorithms (PoW, PoS, DPoS, pBFT, PoA, and Casper) and their limitations, as well as vulnerabilities associated with double-spending and other forms of cyber threats. To enhance resilience and adaptability, a hybridization approach to consensus mechanisms (DPoSW, PoSW, PoCASBFT, DBPoS, etc.) is proposed, integrating the best features of multiple protocols. Additionally, the role of machine learning (ML) methods in dynamic anomaly detection and threat prediction is explored, enabling timely responses to cyberattacks and network parameter optimization. The practical section of the article describes the methodology for deploying an experimental network on the ProximaX platform, which combines decentralized storage, a smart contract layer, and a blockchain ledger. The results of simulating various attacks, including the 51% attack, and analyzing metrics such as network throughput, block confirmation time, and anomaly detection accuracy are presented. Experiments demonstrate that the hybridization of consensus algorithms, combined with ML modules, improves overall security and system scalability, albeit at the cost of additional computational resources. The conclusion highlights future research directions, including the expanded application of reinforcement learning methods, the development of more energy-efficient ML models, and the implementation of advanced privacy-preserving techniques. Approaches to security in distributed applications are of interest to researchers and practitioners specializing in interdisciplinary analyses of information security, distributed computing systems, and cryptographic protocols.