Harnessing dynamic heterogeneous redundancy to empower deep learning safety and security  

在线阅读下载全文

作  者:Fan Zhang Xi Chen Wei Huang Jiangxing Wu Zijie Zhang Chenyu Zhou Jianpeng Li Ziwen Peng Wei Guo Guangze Yang Xinyuan Miao Ruiyang Huang Jiayu Du 

机构地区:[1]National Digital Switching System And Engineering Technological Research Center(NDSC),Zhengzhou 450002,China [2]PLA Information Engineering University,Zhengzhou 450002,China [3]Purple Mountain Laboratories,Nanjing 211111,China [4]Fudan University,Shanghai 200433,China [5]Southeast University,Nanjing 210096,China [6]Zhengzhou University,Zhengzhou 450001,China

出  处:《Security and Safety》2024年第4期86-112,共27页一体化安全(英文)

基  金:supported by the National Key Research and Development Program of China (Project No.2022YFB4500900);the Jiangsu Provincial Department of Science and Technology (Project No. ZL042401)

摘  要:The rapid development of deep learning(DL) models has been accompanied by various safety and security challenges, such as adversarial attacks and backdoor attacks.By analyzing the current literature on attacks and defenses in DL, we find that the ongoing adaptation between attack and defense makes it impossible to completely resolve these issues.In this paper, we propose that this situation is caused by the inherent flaws of DL models,namely non-interpretability, non-recognizability, and non-identifiability. We refer to these issues as the Endogenous Safety and Security(ESS) problems. To mitigate the ESS problems in DL, we propose using the Dynamic Heterogeneous Redundant(DHR) architecture. We believe that introducing diversity is crucial for resolving the ESS problems. To validate the efectiveness of this approach, we conduct various case studies across multiple application domains of DL. Our experimental results confirm that constructing DL systems based on the DHR architecture is more efective than existing DL defense strategies.

关 键 词:Deep learning Endogenous security Dynamic heterogeneous redundancy AI safety 

分 类 号:TP18[自动化与计算机技术—控制理论与控制工程] TP309[自动化与计算机技术—控制科学与工程]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象