摘要
针对网络日志异常信息定位中因网络噪声干扰、多径效应及数据分布不均衡导致异常特征被掩盖、检测准确率低的问题,文章提出一种基于改进YOLOv5s的网络日志异常信息定位方法。首先构建网络信道模型以监测通信节点状态,其次利用改进的YOLOv5s模型提取异常信息特征,并结合分布式定位机制与卡尔曼滤波算法实现异常信息的精确定位。实验结果表明,该方法能有效提升检测准确率,同时降低虚警率。
Aiming at the problem of low detection accuracy in network log anomaly information localization caused by factors such as network noise interference,multipath effects,and imbalanced data distribution,which obscure anomaly features,this article proposes a network log anomaly information localization method based on an improved YOLOv5s model.Firstly,a network channel model is constructed to monitor the states of communication nodes.Subsequently,the improved YOLOv5s model is employed to extract features of anomaly information.Finally,a distributed localization mechanism combined with the Kalman filter algorithm is utilized to achieve precise localization of anomalies.Experimental results demonstrate that the proposed method effectively enhances detection accuracy while reducing the false alarm rate.
作者
张彦玲
ZHANG Yanling(Nanchang Science and Technology Finance Service Center,Nanchang 330038,China)
