为了提高医学诊断模型防御攻击的能力,提出了一种基于生成对抗网络的医学诊断模型知识蒸馏对抗攻击方法。首先创建医学对抗攻击端到端训练网络,并以残差网络作为对抗网络架构;其次在生成器特征块中融合扩张卷积块和通道注意力机制,采用马尔可夫判别器改进判别器网络结构;最后利用生成器和判别器组建生成对抗网络,使用对抗样本进行知识蒸馏对抗攻击,以训练医学诊断模型提高识别精度。采用对抗样本对所提对抗方法进行攻击验证,结果表明:本文方法对抗攻击的成功率为92.6%
与所对比的主流方法相比,该方法的成功率提高了20%
生成对抗样本的最大平均差异降低了3.68%
峰值信噪比、结构相似性分别提升了5.07%、20.29%。本文方法解决了医学诊断模型在对抗攻击中难以获取网络结构和参数信息的问题,生成的对抗样本更接近真实样本,网络效果更佳,为辅助医疗模型诊断及模型安全性提供了参考方案。
于振华,黄山阁,杨波,高红霞,卢思.新型冠状病毒肺炎传播动力学模型构建与分析[J].西安交通大学学报,2022(05).
黄文涵,殷国栋,耿可可,庄伟超,徐利伟.基于扩张卷积特征自适应融合的复杂驾驶场景目标检测[J].东南大学学报(自然科学版),2021(06).
侯静怡,齐雅昀,吴心筱,贾云得.跨语言知识蒸馏的视频中文字幕生成[J].计算机学报,2021(09).
陈法法,成孟腾,杨蕴鹏,陈保家,肖文荣,肖能齐.融合双注意力机制和U-Net网络的锈蚀图像分割[J].西安交通大学学报,2021(12).
李策,李兰,宣树星,杨静,杜少毅.采用超复数小波生成对抗网络的活体人脸检测算法[J].西安交通大学学报,2021(05).
王曙燕,金航,孙家泽.GAN图像对抗样本生成方法[J].计算机科学与探索,2021(04).
张西宁,雷威,杨雨薇,张雯雯.采用自适应基因粒子群算法优化隐马尔科夫模型的方法及应用[J].西安交通大学学报,2018(08).
Rahman Tawsifur,Khandakar Amith,Qiblawey Yazan,Tahir Anas,Kiranyaz Serkan,Abul Kashem Saad Bin,Islam Mohammad Tariqul,Al Maadeed Somaya,Zughaier Susu M.,Khan Muhammad Salman,Chowdhury Muhammad E.H..Exploring the Effect of Image Enhancement Techniques on COVID-19 Detection using Chest X-rays Images[J].Computers in Biology and Medicine,2021.
Hirano Hokuto,Minagi Akinori,Takemoto Kazuhiro.Universal adversarial attacks on deep neural networks for medical image classification[J].BMC Medical Imaging,2021.
Karim Armanious,Chenming Jiang,Marc Fischer,Thomas Küstner,Tobias Hepp,Konstantin Nikolaou,Sergios Gatidis,Bin Yang.MedGAN: Medical image translation using GANs[J].Computerized Medical Imaging and Graphics,2020.
grid.168010.e, 0000000419368956, Department of Electrical Engineering, Stanford University, Stanford, California, USA,grid.168010.e, 0000000419368956, Department of Electrical Engineering, Stanford University, Stanford, California, USA,grid.168010.e, 0000000419368956, Department of Dermatology, Stanford University, Stanford, California, USA,grid.168010.e, 0000000419368956, Department of Pathology, Stanford University, Stanford, California, USA,grid.168010.e, 0000000419368956, Department of Dermatology, Stanford University, Stanford, California, USA,grid.168010.e, 0000000419368956, Department of Dermatology, Stanford University, Stanford, California, USA,grid.280747.e, 0000 0004 0419 2556, Dermatology Service, Veterans Affairs Palo Alto Health Care System, Palo Alto, California, USA,grid.168010.e, 0000000419368956, Department of Microbiology and Immunology, Baxter Laboratory for Stem Cell Biology, Institute for Stem Cell Biology and Regenerative Medicine, Stanford University, Stanford, California, USA,grid.168010.e, 0000000419368956, Department of Computer Science, Stanford University, Stanford, California, USA.Dermatologist-level classification of skin cancer with deep neural networks.[J].Nature,2017.
Varun Gulshan,Lily Peng,Marc Coram,Martin C. Stumpe,Derek Wu,Arunachalam Narayanaswamy,Subhashini Venugopalan,Kasumi Widner,Tom Madams,Jorge Cuadros,Ramasamy Kim,Rajiv Raman,Philip C. Nelson,Jessica L. Mega,Dale R. Webster.Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs[J].JAMA,2016.
Hou Le,Samaras Dimitris,Kurc Tahsin M,Gao Yi,Davis James E,Saltz Joel H.Patch-based Convolutional Neural Network for Whole Slide Tissue Image Classification.[J].Proceedings. IEEE Computer Society Conference on Computer Vision and Pattern Recognition,2016.
Geoffrey E. Hinton,Oriol Vinyals,Jeffrey Dean.Distilling the Knowledge in a Neural Network.[J].CoRR,2015.
0
浏览量
4
下载量
1
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621