基于位图表征与U-Att分类网络的恶意软件识别技术  

Malware Identification Technology Based on Bitmap Representation and U-Att Classification Network

作　　者：屈梦楠 靳宇浩 张光华[1] Qu Mengnan;Jin Yuhao;Zhang Guanghua(School of Information Science and Engineering,Hebei University of Science and Technology,Shijiazhuang 050018)

机构地区：[1]河北科技大学信息科学与工程学院,石家庄050018

出　　处：《信息安全研究》2025年第1期28-34,共7页Journal of Information Security Research

摘　　要：在计算机安全领域,恶意软件识别一直是一个具有挑战性的任务,当前基于深度学习的恶意软件检测技术存在泛化能力不足、性能损耗高等诸多问题.为解决上述问题,提出一种基于位图表征与U-Att分类网络恶意软件识别新技术.U-Att分类网络在残差U-Net网络的基础上,结合了注意力分类器,自适应地聚焦于恶意样本的重要区域,从而提高分类性能.实验中使用多个公开数据集进行了验证,并与其他方法进行了比较分析.实验结果表明,该网络在恶意软件识别任务中取得了优越的性能且拥有更少的参数量.In the field of computer security,malware identification has always been a challenging task.The current malware detection technology based on deep learning has many problems such as insufficient generalization ability and high performance loss.To surmount these obstacles,this paper introduces an innovative technique predicated upon bitmap representation coupled with a U-Att classification network for the discernment of malicious software.This technique augments the residual U-Net architecture with an integrated attention mechanism,culminating in the U-Att classification network that exhibits adaptive focusing on salient regions of malicious samples,thereby ameliorating classification efficacy.Comprehensive validation through the utilization of various public datasets ensued,accompanied by a comparative analysis against alternative methodologies.The empirical findings substantiate the network’s superior performance within the context of malware identification tasks.

关 键 词：恶意软件识别 图像处理 残差U-Net网络 注意力机制 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]