1.Marketing Center of State Grid Sichuan Electric Power Corporation,Chengdu 610045,China
2.College of Computer Science,Sichuan University,Chengdu 610065,China
3.China Electric Power Research Institute,Beijing 102209,China
4.Institute of Computer Science of Sichuan Province,Chengdu 610041,China
Show less
文章历史+
Received
Accepted
Published
2025-05-18
2025-06-03
2026-03-28
Issue Date
2026-04-15
PDF (2399K)
摘要
异常检测旨在识别偏离正常模式的对象,广泛应用于金融、安防、医疗和网络等关键领域。在电力物联网中,异常检测对于应对频谱感知数据篡改攻击尤为重要,有助于保障系统运行稳定,减少潜在经济损失。然而,由于异常样本稀缺、分布不均且模式复杂,传统监督方法难以适应动态变化的实际环境,亟需一种高效且鲁棒的检测方案。为此,本文提出一种基于模糊信息粒球的多尺度异常检测方法(Fuzzy Information Granular-Ball based Outlier Detection,FBOD),通过融合粗粒度与细粒度特征综合评估数据的异常程度。具体工作包含:1) 利用粒球模型自适应划分样本集,并结合密度和邻域信息计算粗粒度异常分数,以捕捉全局异常模式;2) 构建粒球模糊相似距离矩阵,提升对局部异常的识别能力;3) 引入自然邻域异常度量,增强复杂模式的识别效果。实验结果表明,FBOD在多个标准异常检测数据集上AUC均在0.79以上,在电力物联网SSDF攻击数据集上AUC超过0.88,显著优于对比方法。该方法在提升电力系统安全性方面展现出良好适应性与检测性能,同时为其他领域的异常检测任务提供了新思路和技术支持。
Abstract
Anomaly detection aims to identify objects that deviate significantly from normal patterns, and it is widely applied in critical domains such as finance, security, healthcare, and cybersecurity. In the context of the electric Internet of Things (E-IoT), anomaly detection plays a critical role in countering Spectrum Sensing Data Falsification (SSDF) attacks, contributing to system stability and reducing potential economic losses. However, due to the scarcity, imbalance, and complexity of anomaly patterns, traditional supervised methods struggle to adapt to dynamic environments, highlighting the need for a more efficient and robust solution. To address this issue, we propose a multi-scale anomaly detection method based on fuzzy information granules, named FBOD(Fuzzy Information Granular-Ball based Outlier Detection,FBOD). FBOD integrates coarse-grained and fine-grained features to comprehensively assess the degree of anomaly in data. Specifically: 1) a granule ball model adaptively partitions the dataset, and coarse-grained anomaly scores are computed using density and neighborhood information to capture global anomaly patterns; 2) a fuzzy similarity distance matrix is constructed to enhance the detection of local anomalies; and 3) a natural neighborhood-based anomaly metric is introduced to improve the identification of complex patterns. Experimental results show that FBOD achieves AUC scores above 0.79 on multiple benchmark datasets and exceeds 0.88 on an E-IoT dataset under SSDF attacks, significantly outperforming comparative methods. The proposed method demonstrates strong adaptability and detection performance in enhancing the security of power systems, and offers new insights and technical support for anomaly detection in various domains.
异常检测,亦称离群点检测,旨在识别与预期行为或模式显著偏离的对象。在许多数据挖掘研究中,异常点通常被视为噪声而被忽略。然而,在众多应用研究中,识别属于少数类别的离群点可能比识别正常点更重要,例如财务报表欺诈检测[1]、监控[2]、传感器网络[3]和攻击检测[4]。在这些情况下,异常样本的识别是发现潜在安全威胁、性能问题或故障设备的关键步骤。在电力物联网的实际应用中,异常检测在应对频谱感知数据篡改(Spectrum Sensing Data Falsification,SSDF)攻击方面具有重要意义。SSDF攻击是一种复杂的网络攻击,攻击者通过多个节点协同工作,以达到窃取数据、篡改数据或破坏系统正常运行的目的。在电力物联网中,这种攻击可能会导致电力数据的异常波动,从而影响电力系统的稳定运行和电力企业的经济利益。异常检测技术通过实时监测电力数据的异常波动,可以快速识别出潜在的攻击行为,并采取相应的防护措施。这不仅有助于保障电力系统的安全稳定运行,还能减少电力企业的经济损失。因此,异常检测在提升供电可靠性以及保障电力企业经济利益等方面,展现出显著的研究价值和实际应用潜力。
本研究提出了一种基于模糊信息粒球的多尺度异常检测方法(Fuzzy Information Granular-Ball based Outlier Detection,FBOD),从粒球邻域的粗粒度层面和细粒度层面综合度量样本的异常程度。具体实现步骤如下:首先,划分样本集合到一系列粒球集合中,依据每个粒球的密度信息和邻域属性计算每一批样本的粗粒度分数;然后,将粒球模型与模糊信息粒进行深度融合,提出了粒球模糊相似关系,构建了粒球模糊相似距离矩阵;最后结合自然邻域异常分数度量公式,从而得到每个样本的异常分数。本文工作主要包含以下几个方面:
粒球计算是颗粒计算领域发展起来的一种新兴的建模方法,它通过自适应生成的粒球覆盖样本空间并代替原本的数据点输入模式。在无监督学习中,颗粒球计算通常用于各种聚类算法。颗粒球的数量远远小于数据的数量,大大减少了每个聚类算法的运行时间。Cheng等[29]提出了一种基于颗粒球的大规模数据密度峰值聚类(Granular-Ball-based Density Peak clustering,GB-DP)方法。GB-DP采用无监督分区方法创建颗粒球,在不需要参数的情况下,以更短的运行时间获得了类似甚至更好的聚类结果。此外,Cheng等[30]提出了GB-USC,该方法使用颗粒球生成高质量的锚点,用于构建对象和锚点的二部图,以实现低维流形嵌入。GB-USEC通过投票的方式集成了多个GB-USC聚类结果,提高了大规模数据集的运行效率。然而,在该领域的无监督异常点检测方面的研究仍然相对较少[31]。
为了评估本文算法的性能,实验采用AUC(Area Under the Curve)和ROC(Receiver Operating Characteristic Curve)作为评价指标。AUC衡量模型区分不同类别的能力,AUC值越接近1,表示分类性能越优。ROC曲线是分类任务中常用的评估指标,以假阳性率为轴,真阳性率为轴,ROC曲线越接近图像的左上角,表明模型的分类性能越优。
Al-HashediK G, MagalingamP.Financial fraud detection applying data mining techniques: A comprehensive review from 2009 to 2019[J].Comput Sci Rev, 2021, 40: 100402.
[2]
ZhouJ T, DuJ, ZhuH, et al.AnomalyNet: An anomaly detection network for video surveillance[J].IEEE TransInformForensic Secur, 2019, 14(10): 2537-2550.
[3]
PoornimaI G A, ParamasivanB.Anomaly detection in wireless sensor network using machine learning algorithm[J].Comput Commun, 2020, 151: 331-337.
[4]
SiddiquiM A, StokesJ W, SeifertC, et al.Detecting cyber attacks using anomaly detection with explanations and expert feedback[C]//ICASSP 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).Brighton, United Kingdom:IEEE, 2019: 2872-2876.
[5]
XiaS, LiuY, DingX, et al.Granular ball computing classifiers for efficient, scalable and robust learning[J].Inf Sci, 2019, 483: 136-152.
[6]
XieJ, KongW, XiaS, et al.An efficient spectral clustering algorithm based on granular-ball[J].IEEE Trans Knowl Data Eng, 2023, 35(9): 9743-9753.
[7]
XiaS, ZhangH, LiW, et al.GBNRS: A novel rough set algorithm for fast adaptive attribute reduction in classification[J].IEEE Trans Knowl Data Eng, 2022, 34(3): 1231-1242.
[8]
ChenY, WangP, YangX, et al.Granular ball guided selector for attribute reduction[J].Knowl Based Syst, 2021, 229: 107326.
[9]
ZhangK, LengS, PengX, et al.Artificial intelligence inspired transmission scheduling in cognitive vehicular communications and networks[J].IEEE Internet Things J, 2019, 6(2): 1987-1997.
[10]
ChatterjeeP S.Systematic survey on SSDF attack and detection mechanism in cognitive wireless sensor network[C]//2021 International Conference on Intelligent Technologies (CONIT).Hubli, India: IEEE, 2021: 1-5.
[11]
AkyildizI F, LeeW-Y, VuranM C, et al.NeXt generation/dynamic spectrum access/cognitive radio wireless networks: A survey[J].Comput Netw,2006,50(13): 2127-2159.
[12]
ZhengM Q, SunM N, ZengH,et al.Time series anomaly detection of variational autoencoder using fusion diffusion model[J]. Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition),2025, 37(6): 792-804.
YangJ, RahardjaS, FräntiP.Mean-shift outlier detection and filtering[J].Pattern Recognit,2021,115: 107874.
[15]
Santos-FernandezE, Ver HoefJ M, PetersonE E,et al.Unsupervised anomaly detection in spatio-temporal stream network sensor data[J].Water Resour Res,2024,60(11): e2023WR035707.
[16]
ShenH, WeiB, MaY.Unsupervised anomaly detection for manufacturing product images by significant feature space distance measurement[J].Mech Syst Signal Process, 2024,212: 111328.
[17]
GaoC, WangQ, ChenY, et al.A Self-Supervised Contrastive Learning Anomaly Detection Method with Fuzzy Rough Sets[C]//2025 21st International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD).[S.l.]: IEEE, 2025: 435.
[18]
TangB, HeH.A local density-based approach for outlier detection[J].Neurocomputing,2017,241:171-180.
[19]
BaiM, WangX, XinJ,et al.An efficient algorithm for distributed density-based outlier detection on big data[J].Neurocomputing,2016,181: 19-28.
[20]
DegirmenciA, KaralO.Efficient density and cluster based incremental outlier detection in data streams [J].Inf Sci, 2022,607: 901-920.
[21]
YuanZ, ChenB, LiuJ,et al.Anomaly detection based on weighted fuzzy-rough density[J].Appl Soft Comput,2023,134: 109995.
[22]
FangJ, WangZ, LiuW,et al.A new particle swarm optimization algorithm for outlier detection: industrial data clustering in wire arc additive manufacturing [J].IEEE Trans Automat Sci Eng,2024,21(2): 1244-1257.
[23]
JiangS Y, AnQ B.Clustering-based outlier detection method[C]//2008 5th International Conference on Fuzzy Systems and Knowledge Discovery.Jinan Shandong, China: IEEE,2008: 429-433.
[24]
ChristyA, GandhiG M, VaithyasubramanianS.Cluster based outlier detection algorithm for healthcare data[J].Procedia Comput Sci,2015,50: 209-433.
[25]
MalekiS, MalekiS, JenningsN R.Unsupervised anomaly detection with LSTM autoencoders using statistical data-filtering[J].Appl Soft Comput,2021,108: 107443.
[26]
ShylendraA, ShuklaP, MukhopadhyayS,et al.Low power unsupervised anomaly detection by nonparametric modeling of sensor statistics[J].IEEE Trans VLSI Syst, 2020,28(8): 1833-1843.
[27]
YaoR, LiuC, ZhangL,et al.Unsupervised anomaly detection using variational auto-encoder based feature extraction[C]//2019 IEEE International Conference on Prognostics and Health Management (ICPHM).San Francisco,CA,USA:IEEE,2019: 1-7.
[28]
AytekinC, NiX, CricriF,et al.Clustering and unsupervised anomaly detection with l2 normalized deep auto-encoder representations[C]//2018 International Joint Conference on Neural Networks (IJCNN).Rio de Janeiro: IEEE,2018: 1-6.
[29]
FuS, ZhongS, LinL,et al.A re-optimized deep auto-encoder for gas turbine unsupervised anomaly detection[J].Eng Appl Artif Intell,2021,101: 104199.
[30]
ZhouC, PaffenrothR C.Anomaly detection with robust deep autoencoders[C]//Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.Halifax NS Canada:ACM,2017: 665-674.
[31]
ChengD, LiY, XiaS,et al.A fast granular-ball-based density peaks clustering algorithm for large-scale data[J].IEEE Trans Neural Netw Learn Syst,2024,35(12): 17202-17215.
ZhaoX, LiangJ, CaoF.A simple and effective outlier detection algorithm for categorical data[J].Int J Mach Learn Cybern,2014,5(3): 469-477.
[39]
LiuH, LiX, LiJ,et al.Efficient outlier detection for high-dimensional data[J].IEEE Trans Syst Man Cybern, Syst,2018,48(12): 2451-2461.
[40]
XuH, WangY J, WangY J, et al.Mix: A joint learning framework for detecting both clustered and scattered outliers in mixed-type data[C]//2019 IEEE International Conference on Data Mining (ICDM).[S.l.]:IEEE, 2019: 1408-1413.
[41]
LiZ, ZhaoY, BottaN, et al.COPOD: Copula-based outlier detection[C]//2020 IEEE International Conference on Data Mining (ICDM).Sorrento,Italy:IEEE,2020: 1118-1123.
[42]
LiX, LüJ, YiZ.Outlier detection using structural scores in a high-dimensional space[J].IEEE Trans Cybern, 2020,50(5): 2302-2310.
[43]
LiK, GaoX, FuS,et al.Robust outlier detection based on the changing rate of directed density ratio [J].Expert Syst Appl,2022,207: 117988.
[44]
LiZ, ZhaoY, HuX, et al.ECOD: Unsupervised outlier detection using empirical cumulative distribution functions[J].IEEE Trans Knowl Data Eng,2023,35(12): 12181-12193.
[45]
YuanZ, ChenB, LiuJ,et al.Anomaly detection based on weighted fuzzy-rough density[J].Appl Soft Comput, 2023,134: 109995.
[46]
YangJ, RahardjaS, FräntiP.Mean-shift outlier detection and filtering[J].Pattern Recognit,2021,115: 107874.
[47]
XuH, PangG, WangY,et al.Deep isolation forest for anomaly detection[J].IEEE Trans Knowl Data Eng,2023, 35(12): 12591-12604.
[48]
RuffL, VandermeulenR, GoernitzN,et al.Deep one-class classification[C]//International Conference on Machine Learning. Stockholm SWEDEN: PMLR,2018: 4393-4402.
[49]
FriedmanM.A comparison of alternative tests of significance for the problem of m rankings[J].Ann Math Statist,1940,11(1): 86-92.
[50]
LiX H, YuZ N, ShaoS S, et al. Synergistic effect and kinetics analysis of rape straw and waste plastics co-catalytic pyrolysis[J]. Journal of Jiangsu University(Natural Science Edition),2025, 46(2): 156-162.