Page 406 - 《软件学报》2025年第5期
P. 406
2306 软件学报 2025 年第 36 卷第 5 期
神经网络的节点分类任务. 首先, 针对数据不一致问题以及数据不平衡问题, 提出基于类别感知的恶意节点检测方
法 CAMD. CAMD 通过引入类别感知注意力系数、不一致图神经网络编码器以及类别感知不平衡损失函数, 增加
了不同类型节点表示的区分度, 保留了节点原本的信息以及表示学习过程中不同局域性的信息, 增加节点表示的
+
表达能力. 接下来, 针对 CAMD 在标签稀缺情况下效果受限的问题, 提出了基于图对比学习的方法 CAMD , 引入
数据增强、自监督图对比学习以及基于类别感知的平衡图对比学习, 使模型在标签稀缺情况下取得良好的效果.
+
在 5 个真实世界数据集上的大量实验表明, 本文提出的 CAMD 与 CAMD 方法的恶意节点检测性能由于其他基线
方法. 未来的工作包括进一步探索数据不一致场景下, 基于图的标签传播方法对模型的效果提升, 以及考虑采用类
似 GraphSAGE 的归纳式学习方法以减少计算内存占用, 从而使模型适用于更大规模的图数据.
References:
[1] Liu ZQ, Chen CC, Yang XX, Zhou J, Li XL, Song L. Heterogeneous graph neural networks for malicious account detection. In: Proc. of
the 27th ACM Int’l Conf. on Information and Knowledge Management. Torino: ACM, 2018. 2077–2085. [doi: 10.1145/3269206.
3272010]
[2] Liu ZW, Dou YT, Yu PS, Deng YT, Peng H. Alleviating the inconsistency problem of applying graph neural network to fraud detection.
In: Proc. of the 43rd Int’l ACM SIGIR Conf. on Research and Development in Information Retrieval. ACM, 2020. 1569–1572. [doi: 10.
1145/3397271.3401253]
[3] Liang C, Liu ZQ, Liu B, Zhou J, Li XL, Yang S, Qi Y. Uncovering insurance fraud conspiracy with network learning. In: Proc. of the
42nd Int’l ACM SIGIR Conf. on Research and Development in Information Retrieval. Paris: ACM, 2019. 1181–1184. [doi: 10.1145/
3331184.3331372]
[4] Liu ZQ, Chen CC, Li LF, Zhou J, Li XL, Song L, Qi Y. GeniePath: Graph neural networks with adaptive receptive paths. In: Proc. of the
33rd AAAI Conf. on Artificial Intelligence. Honolulu: AAAI, 2019. 4424–4431. [doi: 10.1609/aaai.v33i01.33014424]
[5] Wang DX, Lin JB, Cui P, Jia QH, Wang Z, Fang YM, Yu Q, Zhou J, Yang S, Qi Y. A semi-supervised graph attentive network for
financial fraud detection. In: Proc. of the 2019 IEEE Int’l Conf. on Data Mining (ICDM). Beijing: IEEE, 2019. 598–607. [doi: 10.1109/
ICDM.2019.00070]
[6] Kumar S, Zhang XK, Leskovec J. Predicting dynamic embedding trajectory in temporal interaction networks. In: Proc. of the 25th ACM
SIGKDD Int’l Conf. on Knowledge Discovery & Data Mining. Anchorage: ACM, 2019. 1269–1278. [doi: 10.1145/3292500.3330895]
[7] Dou YT, Liu ZW, Sun L, Deng YT, Peng H, Yu PS. Enhancing graph neural network-based fraud detectors against camouflaged
fraudsters. In: Proc. of the 29th ACM Int’l Conf. on Information & Knowledge Management. ACM, 2020. 315–324. [doi: 10.1145/
3340531.3411903]
[8] Liu Y, Ao X, Qin ZD, Chi JF, Feng JH, Yang H, He Q. Pick and choose: A GNN-based imbalanced learning approach for fraud
detection. In: Proc. of the 2021 Web Conf. Ljubljana: ACM, 2021. 3168–3177. [doi: 10.1145/3442381.3449989]
[9] Wang YL, Zhang J, Guo SS, Yin HZ, Li CP, Chen H. Decoupling representation learning and classification for GNN-based anomaly
detection. In: Proc. of the 44th Int’l ACM SIGIR Conf. on Research and Development in Information Retrieval. ACM, 2021. 1239–1248.
[doi: 10.1145/3404835.3462944]
[10] Veličković P, Fedus W, Hamilton WL, Liò P, Bengio Y, Hjelm RD. Deep graph infomax. In: Proc. of the 2019 Int’l Conf. on Learning
Representations. New Orleans: OpenReview.net, 2019.
[11] Chen B, Zhang J, Zhang XK, Dong YX, Song J, Zhang P, Xu KB, Kharlamov E, Tang J. GCCAD: Graph contrastive coding for anomaly
detection. IEEE Trans. on Knowledge and Data Engineering, 2023, 35(8): 8037–8051. [doi: 10.1109/TKDE.2022.3200459]
[12] Tang JH, Li JJ, Gao ZQ, Li J. Rethinking graph neural networks for anomaly detection. In: Proc. of the 39th Int’l Conf. on Machine
Learning. Baltimore: PMLR, 2022. 21076–21089.
[13] Chai ZW, You SQ, Yang Y, Pu SL, Xu JR, Cai HY, Jiang WH. Can abnormality be detected by graph neural networks? In: Proc. of the
31st Int’l Joint Conf. on Artificial Intelligence. Vienna, 2022. 1945–1951. [doi: 10.24963/ijcai.2022/267]
[14] Shi FZ, Cao YN, Shang YM, Zhou YC, Zhou C, Wu J. H2-FDetector: A GNN-based fraud detector with homophilic and heterophilic
connections. In: Proc. of the 2022 ACM Web Conf. ACM, 2022. 1486–1494. [doi: 10.1145/3485447.3512195]
[15] Lim D, Hohne F, Li XY, Huang SL, Gupta V, Bhalerao O, Lim SN. Large scale learning on non-homophilous graphs: New benchmarks
and strong simple methods. In: Proc. of the 35th Conf. on Neural Information Processing Systems (NeurIPS 2021). Montreal, 2021. 1–16.
[16] Abu-El-Haija S, Perozzi B, Kapoor A, Alipourfard N, Lerman K, Harutyunyan H, Steeg GV, Galstyan A. MixHop: Higher-order graph
convolutional architectures via sparsified neighborhood mixing. In: Proc. of the 36th Int’l Conf. on Machine Learning. Long Beach: