高判别精度的区块链交易合法性检测方法

doi:10.3778/j.issn.1002-8331.2211-0395

摘要/Abstract

摘要： 区块链上的交易合法性检测对于加密数字货币的监管具有重大意义。针对现有交易合法性检测方法存在的检测精度低下、判别过程中难以有效兼顾交易本身信息与前后拓扑信息的问题，提出融合可信深度森林的多角度高精度合法性检测方法。设计基于可信生成特征的可信深度森林TForest，以特征重排序的方式赋予子样本足够的区分度，结合可变滑动窗口以均衡无混淆的方式提取可信子样本，在大幅度降低生成特征维度的基础上，提高了深度森林的判别精度。提出一种集成策略，基于不同基模型对于正负样本识别能力的差异性，采用双阶段逐层优化的方式有效融合可信深度森林与Transformer图网络及残差网络三类基判别器，兼顾两方面信息，构成高精度的多角度分析模型T2Rnet。在Elliptic数据集上的实验结果显示，该模型的F1-score达到83.11%，相比基准图卷积方法提升31.6%，具备可靠的交易合法性检测性能。

关键词: 区块链, 合法性检测, 可信深度森林, 神经网络, 双阶段集成

Abstract: Legitimacy discrimination of transactions on the blockchain is of great importance for the regulation of cryptocurrencies. In order to effectively take into account the information of the transaction itself and the topological information in the discriminative process, and to improve the discrimination accuracy, this paper proposes a multi-perspective legitimacy detection method that incorporates the trustworthy deep forest. Firstly, a trustworthy deep forest (TForest) based on generating trustworthy features is designed. It gives sufficient discrimination to subsamples by feature reordering and combines variable sliding windows to extract differentiable subsamples in a balanced and confusion-free manner. The discrimination accuracy of the deep forest is improved on the basis of significantly reducing the dimensionality of generated features. Then, an ensemble strategy is designed. It uses a two-stage layer-by-layer optimization approach to effectively fuse three types of base discriminators, namely trustworthy deep forest, Transformer graph network and ResNet. The strategy is based on the difference of base models for positive and negative samples recognition ability, and utilizes two kinds of information, finally, a high-accuracy multi-perspective analysis model T2Rnet is constituted. The experimental results on the Elliptic dataset show that the F1-score of the model achieves 83.11%, which is 31.6% higher than the baseline graph convolution method. The model has reliable transaction legitimacy discrimination performance.

Key words: blockchain, legitimacy discrimination, trustworthy deep forest, neural network, two-stage ensemble strategy

蔡元海, 宋甫元, 黎凯, 陈彦宇, 付章杰. 高判别精度的区块链交易合法性检测方法[J]. 计算机工程与应用, 2024, 60(5): 271-280.

CAI Yuanhai, SONG Fuyuan, LI Kai, CHEN Yanyu, FU Zhangjie. Blockchain Transaction Legitimacy Discrimination with High Recognition Accuracy[J]. Computer Engineering and Applications, 2024, 60(5): 271-280.

参考文献

[1] MCGINN D, BIRCH D, AKROYD D, et al. Visualizing dynamic Bitcoin transaction patterns[J]. Big Data, 2016, 4(2): 109-119.
[2] BISTARELLI S, SANTINI F. Go with the-bitcoin-flow, with visual analytics[C]//Proceedings of the 12th International Conference on Availability, Reliability and Security, 2017: 1-6.
[3] BATTISTA D, DONATO D, PATRIGNANI M, et al. Bitconeview: visualization of flows in the bitcoin transaction graph[C]//Proceedings of the IEEE Symposium on Visualization for Cyber Security, 2015: 1-8.
[4] CONTI M, GANGWAL A, RUJ S. On the economic significance of ransomware campaigns: a Bitcoin transactions perspective[J]. Computers & Security, 2018, 79: 162-189.
[5] PHAM T, LEE S. Anomaly detection in bitcoin network using unsupervised learning methods[J]. arXiv:1611.03941,2016.
[6] 沈蒙, 桑安琪, 祝烈煌, 等. 基于动机分析的区块链数字货币异常交易行为识别方法[J]. 计算机学报, 2021, 44(1): 193-208.
SHEN M, SANG A Q, ZHU L H, et al. Abnormal transaction behavior recognition based on motivation analysis in Blockchain digital currency[J]. Chinese Journal of Computers, 2021, 44(1): 193-208.
[7] HARLEV M A, SUN Y H, LANGENHELDT K C, et al. Breaking bad: de-anonymising entity types on the bitcoin blockchain using supervised machine learning[C]//Proceedings of the 51st Hawaii International Conference on System Sciences, 2018.
[8] WEBER M, DOMENICONI G, CHEN J, et al. Anti-money laundering in Bitcoin: experimenting with graph convolutional networks for financial forensics[C]//Proceedings of the SIGKDD Workshop on Anomaly Detection in Finance, Anchorage, AK, 2019.
[9] 朱会娟, 陈锦富, 李致远, 等. 基于多特征自适应融合的区块链异常交易检测方法[J]. 通信学报, 2021, 42(5): 41-50.
ZHU H J, CHEN J F, LI Z Y, et al. Block-chain abnormal transaction detection method based on adaptive multi-feature fusion[J]. Journal on Communications, 2021, 42(5): 41-50.
[10] CHAI Z, YOU S, YANG Y, et al. Can abnormality be detected by graph neural networks?[C]//Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, Vienna, July, 2022: 1945-1951.
[11] PAREJA A, DOMENICONI G, CHEN J, et al. EvolveGCN: evolving graph convolutional networks for dynamic graphs[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 5363-5370.
[12] ZHOU Z H, FENG J. Deep forest[J]. National Science Review, 2019, 6(1): 74-86.
[13] WELLING M, KIPF T N. Semi-supervised classification with graph convolutional networks[C]//Proceedings of the International Conference on Learning Representations, 2016.
[14] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Advances in Neural Information Processing Systems, 2017: 5998-6008.
[15] SHI Y, HUANG Z, FENG S, et al. Masked label prediction: unified message passing model for semi-supervised classification[J]. arXiv:2009.03509,2020.
[16] HE K M, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, 2016: 770-778.
[17] CHEN T, GUESTRIN C. XGBoost: a scalable tree boosting system[C]//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016: 785-794.
[18] MAATEN L, HINTON G. Visualizing data using t-SNE[J]. Journal of Machine Learning Research, 2008, 9: 2579-2605.
[19] VELICKOVIC P, CUCURULL G, Casanova A, et al. Graph attention networks[C]//ICLR 2018 Conference Blind Submission, 2018.
[20] KE G, MENG Q, FINLEY T, et al. LightGBM: a highly efficient gradient boosting decision tree[C]//Advances in Neural Information Processing Systems, 2017: 3146-3154.