To address the issue of category concept drift in network protocol data, a ViT-KANs-based dual-head algorithm for communication network protocol data category concept drift detection and classification is proposed. The global perception capability of vision transformer (ViT) and the flexible function approximation ability of kolmogorov-arnold networks (KANs) are integrated to construct an efficient feature extraction network in this algorithm. A dual-head parallel output structure is adopted to handle the classification of old-class data and the detection of category concept drift, respectively. Furthermore, the validation set data is utilized to adaptively compute the confidence threshold, which effectively alleviates the lack of concept drift samples during the training phase. Experiments are conducted on three datasets, namely the Moore dataset, the Canadian Institute for Cybersecurity Intrusion Detection Evaluation Dataset 2017 (CICIDS2017), and the improved version of the Network Security Lab (NSL)-Knowledge Discovery and Data Mining Competition Dataset (NSL-KDD). The results show that the detection error rate of the proposed method is reduced significantly, compared to those of models and out-of-distribution detection methods, while superior classification accuracy category is maintained.
针对上述问题,提出一种基于ViT-KANs的双头通信网络数据协议类别概念漂移检测分类算法(A ViT-KANs-based Dual-Head Algorithm for Communication Network Protocol Data Category Concept Drift Detection and Classification, ViTKANs-DH)。首先,设计Transformer-KANs混合编码器,融合注意力与样条函数以增强特征表征;其次,构建双头并行输出结构,通过联合学习同步实现分类与漂移检测;最后,提出基于验证集的置信度阈值确定方法,为漂移识别提供依据。
采用3个广泛认可的数据集来验证所提算法的有效性,包括Moore数据集、CICIDS2017数据集和NSL-KDD数据集。选择这3个数据集的目的是覆盖不同类型的网络流量特征,从而全面评估算法的泛化能力。Moore数据集由剑桥大学采集,是网络流量分析领域的经典数据集;CICIDS2017数据集是由加拿大网络安全研究所采集并发布的网络安全数据集;NSL-KDD数据集是KDD Cup 1999数据集的改进版本,广泛应用于网络数据分析领域。
AGRAHARIS, SINGHA K. Concept drift detection in data stream mining: a literature review[J]. Journal of King Saud University-Computer and Information Sciences, 2022,34(10):9523-9540.
[2]
AGRAHARIS, SINGHA K. Disposition-based concept drift detection and adaptation in data stream[J]. Arabian Journal for Science and Engineering, 2022,47(8):10605-10621.
[3]
LIUA J, SONGY L, ZHANGG Q, et al. Regional concept drift detection and density synchronized drift adaptation[C]∥Proceedings of the 26th International Joint Conference on Artificial Intelligence. New York, USA: ACM, 2017:2280-2286.
[4]
SCHLIMMERJ C, GRANGERR H. Incremental learning from noisy data[J]. Machine Learning, 1986,1(3):317-354.
[5]
BAZZANA L C, LABIDIS. Advances in artificial intelligence-SBIA 2004: 17th Brazilian symposium on artificial intelligence proceedings[M]. Cham, Switzerland: Springer, 2004:286-295.
[6]
BAENA-GARCÍAM, CAMPO-ÁVILAJ D, FIDALGOR, et al. Early drift detection method [C]∥Proceedings of the 4th International Workshop on Knowledge Discovery from Data Streams. Cham, Switzerland: Springer, 2006,6:77-86.
[7]
BIFETA, GAVALDÀR. Learning from time-changing data with adaptive windowing[C]∥Proceedings of the 2007 SIAM International Conference on Data Mining. Philadelphia, USA: Society for Industrial and Applied Mathematics, 2007:443-448.
[8]
HAQUEA, KHANL, BARONM. SAND: semi-supervised adaptive novel class detection and classification over data stream[C]∥Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence. Palo Alto, USA: AAAI Press, 2016:1652-1658.
[9]
HENDRYCKSD, GIMPELK. A baseline for detecting misclassified and out-of-distribution examples in neural networks[DB/OL]. (2016-10-07)[2025-02-01].
[10]
LIANGS, LIY, SRIKANTR. Enhancing the reliability of out-of-distribution image detection in neural networks[DB/OL]. (2017-06-08)[2025-02-01].
[11]
ASHFAHANIA, PRATAMAM, LUGHOFERE, et al. DEVDAN: deep evolving denoising autoencoder[J]. Neurocomputing, 2020,390:297-314.
[12]
LUOX, ZHANGD X. An adaptive deep learning framework for day-ahead forecasting of photovoltaic power generation[J]. Sustainable Energy Technologies and Assessments, 2022,52:No.102326.
[13]
GUANH T, WANGY J, MAX K, et al. DCIGAN: a distributed class-incremental learning method based on generative adversarial networks[C]∥Proceedings of the 2019 IEEE International Conference on Parallel & Distributed Processing with Applications, Big Data & Cloud Computing, Sustainable Computing & Communications, Social Computing & Networking. Piscataway, USA: IEEE, 2019:768-775.
[14]
XUL J, DINGX, PENGH P, et al. ADTCD: an adaptive anomaly detection approach toward concept drift in IoT[J]. IEEE Internet of Things Journal, 2023,10(18):15931-15942.
[15]
DOSOVITSKIYA, BEYERL, KOLESNIKOVA, et al. An image is worth 16x16 words: transformers for image recognition at scale[DB/OL]. (2020-10-22)[2025-02-01].
[16]
VASWANIA, SHAZEERN M, PARMARN, et al. Attention is all you need[C]∥Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook, USA: Curran Associates Inc., 2017:6000-6010.
[17]
HOC M K, YOWK C, ZHUZ W, et al. Network intrusion detection via flow-to-image conversion and vision transformer classification[J]. IEEE Access, 2022,10:97780-97793.
[18]
WASSWAH, LYNART, NANYONGAA, et al. IoT botnet detection: application of vision transformer to classification of network flow traffic[C]∥Proceedings of the 2023 Global Conference on Information Technologies and Communications. Piscataway, USA: IEEE, 2023:1-6.
[19]
LIUZ, WANGY, VAIDYAS, et al. KAN: Kolmogorov-Arnold networks[DB/OL]. (2024-04-30)[2025-02-01].
[20]
GAOW, GONGZ, DENGZ, et al. TabKANet: tabular data modeling with Kolmogorov-Arnold network and transformer[DB/OL]. (2024-09-13)[2025-02-01].
[21]
GOODFELLOWI J, SHLENSJ, SZEGEDYC. Explaining and harnessing adversarial examples[DB/OL]. (2014-12-20)[2025-02-01].
[22]
CHAWLAN V, BOWYERK W, HALLL O, et al. SMOTE: synthetic minority over-sampling technique[J]. Journal of Artificial Intelligence Research, 2002,16(1):321-357.
HANH, WANGW Y, MAOB H. Borderline-SMOTE: a new over-sampling method in imbalanced data sets learning[C]∥Proceedings of the Advances in Intelligent Computing. Cham, Switzerland: Springer, 2005:878-887.
[25]
HEH B, BAIY, GARCIAE A, et al. ADASYN: adaptive synthetic sampling approach for imbalanced learning[C]∥Proceedings of the 2008 IEEE International Joint Conference on Neural Networks. Piscataway, USA: IEEE, 2008:1322-1328.