Anti-D chain: A lightweight DDoS attack detection scheme based on heterogeneous ensemble learning in blockchain
Bin Jia, Yongquan Liang
Abstract
With rapid development of blockchain technology, blockchain and its security theory research and practical application have become crucial. At present, a new DDoS attack has arisen, and it is the DDoS attack in blockchain network. The attack is harmful for blockchain technology and many application scenarios. However, the traditional and existing DDoS attack detection and defense means mainly come from the centralized tactics and solution. Aiming at the above problem, the paper proposes the virtual reality parallel an-ti-DDoS chain design philosophy and distributed anti-D Chain detection framework based on hybrid ensemble learning. Here, AdaBoost and Random Forest are used as our ensemble learning strategy, and some different lightweight classifiers are integrated into the same ensemble learning algorithm, such as CART and ID3. Our detection framework in block-chain scene has much stronger generalization performance, universality and complementarity to identify accurately the onslaught features for DDoS attack in P2P network. Extensive experimental results confirm that our distributed heterogeneous anti-D chain detection method has better performance in six important indicators (such as Precision, Recall, F-Score, True Positive Rate, False Positive Rate, and ROC curve).