Adversarial distillation(AD)has emerged as a potential solution to tackle the challenging optimization problem of loss with hard labels in adversarial training.However,fixed sample-agnostic and student-egocentric atta...Adversarial distillation(AD)has emerged as a potential solution to tackle the challenging optimization problem of loss with hard labels in adversarial training.However,fixed sample-agnostic and student-egocentric attack strategies are unsuitable for distillation.Additionally,the reliability of guidance from static teachers diminishes as target models become more robust.This paper proposes an AD method called Learnable Distillation Attack Strategies and Evolvable Teachers Adversarial Distillation(LDAS&ET-AD).Firstly,a learnable distillation attack strategies generating mechanism is developed to automatically generate sample-dependent attack strategies tailored for distillation.A strategy model is introduced to produce attack strategies that enable adversarial examples(AEs)to be created in areas where the target model significantly diverges from the teachers by competing with the target model in minimizing or maximizing the AD loss.Secondly,a teacher evolution strategy is introduced to enhance the reliability and effectiveness of knowledge in improving the generalization performance of the target model.By calculating the experimentally updated target model’s validation performance on both clean samples and AEs,the impact of distillation from each training sample and AE on the target model’s generalization and robustness abilities is assessed to serve as feedback to fine-tune standard and robust teachers accordingly.Experiments evaluate the performance of LDAS&ET-AD against different adversarial attacks on the CIFAR-10 and CIFAR-100 datasets.The experimental results demonstrate that the proposed method achieves a robust precision of 45.39%and 42.63%against AutoAttack(AA)on the CIFAR-10 dataset for ResNet-18 and MobileNet-V2,respectively,marking an improvement of 2.31%and 3.49%over the baseline method.In comparison to state-of-the-art adversarial defense techniques,our method surpasses Introspective Adversarial Distillation,the top-performing method in terms of robustness under AA attack for the CIFAR-10 dataset,with enhancements of 1.40%and 1.43%for ResNet-18 and MobileNet-V2,respectively.These findings demonstrate the effectiveness of our proposed method in enhancing the robustness of deep learning networks(DNNs)against prevalent adversarial attacks when compared to other competing methods.In conclusion,LDAS&ET-AD provides reliable and informative soft labels to one of the most promising defense methods,AT,alleviating the limitations of untrusted teachers and unsuitable AEs in existing AD techniques.We hope this paper promotes the development of DNNs in real-world trust-sensitive fields and helps ensure a more secure and dependable future for artificial intelligence systems.展开更多
Uncertain security threats caused by vulnerabilities and backdoors are the most serious and difficult problem in cyberspace.This paper analyzes the philosophical and technical causes of the existence of so-called"...Uncertain security threats caused by vulnerabilities and backdoors are the most serious and difficult problem in cyberspace.This paper analyzes the philosophical and technical causes of the existence of so-called"dark functions"such as system vulnerabilities and backdoors,and points out that endogenous security problems cannot be completely eliminated at the theoretical and engineering levels;rather,it is necessary to develop or utilize the endogenous security functions of the system architecture itself.In addition,this paper gives a definition for and lists the main technical characteristics of endogenous safety and security in cyberspace,introduces endogenous safety and security mechanisms and characteristics based on dynamic heterogeneous redundancy(DHR)architecture,and describes the theoretical implications of a coding channel based on DHR.展开更多
Elastic control could balance the distributed control plane in Software-Defined Networking(SDN). Dynamic switch migration has been proposed to achieve it. However, existing schemes mainly focus on how to execute migra...Elastic control could balance the distributed control plane in Software-Defined Networking(SDN). Dynamic switch migration has been proposed to achieve it. However, existing schemes mainly focus on how to execute migration operation, but not why. This paper designs a decision-making mechanism based on zero-sum game theory to reelect a new controller as the master for migrated switches. It first chooses a switch for migration in the heavy controller which invites its neighbors as the game players to compete for the master role of this switch in the game-playing field(GPF) which is an occasional and loose domain for game-playing. Second, based on the concept of GPF, we design a decentralized strategy to play the game and determine which player as the final master. We implement it by extending the Open Flow protocol. Finally, numerical results demonstrate that our distributed strategy can approach elastic control plane with better performance.展开更多
Shortest-path calculation on weighted graphs are an essential operation in computer networks. The performance of such algorithms has become a critical challenge in emerging software-defined networks(SDN),since SDN con...Shortest-path calculation on weighted graphs are an essential operation in computer networks. The performance of such algorithms has become a critical challenge in emerging software-defined networks(SDN),since SDN controllers need to centralizedly perform a shortest-path query for every flow,usually on large-scale network. Unfortunately,one of the challenges is that current algorithms will become incalculable as the network size increases. Therefore, inspired by the compression graph in the field of compute visualization,we propose an efficient shortest path algorithm by compressing the original big network graph into a small one, but the important graph properties used to calculate path is reserved. We implement a centralized version of our approach in SDN-enabled network,and the evaluations validate the improvement compared with the well-known algorithms.展开更多
This special topic mainly focuses on the progress of physical-layer security(PLS)technologies and their potential applications for the future beyond fifth-generation(B5G)and sixth-generation(6G)networks.The tremendous...This special topic mainly focuses on the progress of physical-layer security(PLS)technologies and their potential applications for the future beyond fifth-generation(B5G)and sixth-generation(6G)networks.The tremendous growth in connectivity and the ubiquity of wireless communications have resulted in an unprecedented awareness of the importance of security and privacy.Achieving secure and trusted communications is vital for future intelligent connected applications,especially life-critical vehicle-toeverything(V2X)applications.However,the heterogeneous,dynamic and decentralized architecture of these networks leads to difficulties for cryptographic key management,and distribution.By exploiting the physical characteristics of devices,wireless channels and noise,PLS offers reliable solutions against eavesdropper attacks as complementary approaches to cryptographic techniques.展开更多
With the growing amount of information and data, object-oriented storage systems have been widely used in many applications, including the Google File System, Amazon S3, Hadoop Distributed File System, and Ceph, in wh...With the growing amount of information and data, object-oriented storage systems have been widely used in many applications, including the Google File System, Amazon S3, Hadoop Distributed File System, and Ceph, in which load balancing of metadata plays an important role in improving the input/output performance of the entire system. Unbalanced load on the metadata server leads to a serious bottleneck problem for system performance. However, most existing metadata load balancing strategies, which are based on subtree segmentation or hashing, lack good dynamics and adaptability. In this study, we propose a metadata dynamic load balancing(MDLB) mechanism based on reinforcement learning(RL). We learn that the Q_learning algorithm and our RL-based strategy consist of three modules, i.e., the policy selection network, load balancing network, and parameter update network. Experimental results show that the proposed MDLB algorithm can adjust the load dynamically according to the performance of the metadata servers, and that it has good adaptability in the case of sudden change of data volume.展开更多
The sixth-generation mobile communication(6G)networks will face more complex endogenous security problems,and it is urgent to propose new universal security theories and establish new practice norms to deal with the&#...The sixth-generation mobile communication(6G)networks will face more complex endogenous security problems,and it is urgent to propose new universal security theories and establish new practice norms to deal with theªunknown unknownºsecurity threats in cyberspace.This paper first expounds the new paradigm of cyberspace endogenous security and introduces the vision of 6G cyberspace security.Then,it analyzes the security problems faced by the 6G core network,wireless access network,and emerging associated technologies in detail,as well as the corresponding security technology development status and the integrated development of endogenous security and traditional security.Furthermore,this paper describes the relevant security theories and technical concepts under the guidance of the new paradigm of endogenous security.展开更多
The common endogenous security problems in cyberspace and related attack threats have posed subversive challenges to conventional theories and methods of functional safety.In the current design of the cyber physical s...The common endogenous security problems in cyberspace and related attack threats have posed subversive challenges to conventional theories and methods of functional safety.In the current design of the cyber physical system(CPS),functional safety and cyber security are increasingly intertwined and inseparable,which evolve into the generalized functional safety(S&S)problem.The conventional reliability and cybersecurity technologies are unable to provide security assurance with quanti able design and veri cation metrics in response to the cyberattacks in hardware and software with common endogenous security problems,and the functional safety of CPS facilities or device has become a frightening ghost.The dynamic heterogeneity redundancy(DHR)architecture and coding channel theory(CCT)proposed by the cyberspace endogenous security paradigm could handle random failures and uncertain network attacks in an integrated manner,and its generalized robust control mechanism can solve the universal problem of quantitative design for functional safety under probability or improbability perturbation.As a generalized functional safety enabling structure,DHR opens up a new direction to solve the common endogenous security problems in the cross-disciplinary elds of cyberspace.展开更多
Recent advances in deep learning have led to disruptive breakthroughs in artificial intelligence(AI),fueling the jump in ChatGPT-like large language models(LLMs).As with any emerging technology,it is a two-sided coin,...Recent advances in deep learning have led to disruptive breakthroughs in artificial intelligence(AI),fueling the jump in ChatGPT-like large language models(LLMs).As with any emerging technology,it is a two-sided coin,bringing not only vast social impacts but also significant security concerns,especially in the socio-cognitive domain.Against this back-ground,this work starts with an inherent mechanism analysis of cognitive domain games,from which it proceeds to explore the security concerns facing the cognitive domain as well as to analyze the formation mechanisms of a cognitive immune system.Finally,inspired by behavioral mimicry in biology,this work will elaborate on new approaches to cognitive security from three aspects:Mimicry Computing,Mimicry Defense,and Mimicry Intelligence.展开更多
Aiming at the problem of failure recovery in current networks,a fast failure recovery method based on equivalent cooperative routing is proposed.Firstly,the transmission path between the source and destination nodes i...Aiming at the problem of failure recovery in current networks,a fast failure recovery method based on equivalent cooperative routing is proposed.Firstly,the transmission path between the source and destination nodes is divided into several non-overlapping path segments.Next,backup paths are deployed for each link in the path segment through segmented routing technology,which ensures fast routing recovery after failure.Additionally,in order to avoid damaging the QoS of the data stream through the failure recovery process,the transmission is guaranteed by the intersegment QoS complement.The experimental results show that the proposed method has a low failure recovery delay under a relatively small flow table cost.展开更多
基金the National Key Research and Development Program of China(2021YFB1006200)Major Science and Technology Project of Henan Province in China(221100211200).Grant was received by S.Li.
文摘Adversarial distillation(AD)has emerged as a potential solution to tackle the challenging optimization problem of loss with hard labels in adversarial training.However,fixed sample-agnostic and student-egocentric attack strategies are unsuitable for distillation.Additionally,the reliability of guidance from static teachers diminishes as target models become more robust.This paper proposes an AD method called Learnable Distillation Attack Strategies and Evolvable Teachers Adversarial Distillation(LDAS&ET-AD).Firstly,a learnable distillation attack strategies generating mechanism is developed to automatically generate sample-dependent attack strategies tailored for distillation.A strategy model is introduced to produce attack strategies that enable adversarial examples(AEs)to be created in areas where the target model significantly diverges from the teachers by competing with the target model in minimizing or maximizing the AD loss.Secondly,a teacher evolution strategy is introduced to enhance the reliability and effectiveness of knowledge in improving the generalization performance of the target model.By calculating the experimentally updated target model’s validation performance on both clean samples and AEs,the impact of distillation from each training sample and AE on the target model’s generalization and robustness abilities is assessed to serve as feedback to fine-tune standard and robust teachers accordingly.Experiments evaluate the performance of LDAS&ET-AD against different adversarial attacks on the CIFAR-10 and CIFAR-100 datasets.The experimental results demonstrate that the proposed method achieves a robust precision of 45.39%and 42.63%against AutoAttack(AA)on the CIFAR-10 dataset for ResNet-18 and MobileNet-V2,respectively,marking an improvement of 2.31%and 3.49%over the baseline method.In comparison to state-of-the-art adversarial defense techniques,our method surpasses Introspective Adversarial Distillation,the top-performing method in terms of robustness under AA attack for the CIFAR-10 dataset,with enhancements of 1.40%and 1.43%for ResNet-18 and MobileNet-V2,respectively.These findings demonstrate the effectiveness of our proposed method in enhancing the robustness of deep learning networks(DNNs)against prevalent adversarial attacks when compared to other competing methods.In conclusion,LDAS&ET-AD provides reliable and informative soft labels to one of the most promising defense methods,AT,alleviating the limitations of untrusted teachers and unsuitable AEs in existing AD techniques.We hope this paper promotes the development of DNNs in real-world trust-sensitive fields and helps ensure a more secure and dependable future for artificial intelligence systems.
基金supported by the National Natural Science Foundation Innovation Group Project(61521003)。
文摘Uncertain security threats caused by vulnerabilities and backdoors are the most serious and difficult problem in cyberspace.This paper analyzes the philosophical and technical causes of the existence of so-called"dark functions"such as system vulnerabilities and backdoors,and points out that endogenous security problems cannot be completely eliminated at the theoretical and engineering levels;rather,it is necessary to develop or utilize the endogenous security functions of the system architecture itself.In addition,this paper gives a definition for and lists the main technical characteristics of endogenous safety and security in cyberspace,introduces endogenous safety and security mechanisms and characteristics based on dynamic heterogeneous redundancy(DHR)architecture,and describes the theoretical implications of a coding channel based on DHR.
基金supported by the Foundation for Innovative Research Groups of the National Natural Science Foundation of China(Grant No.61521003)the National Basic Research Program of China(2012CB315901,2013CB329104)+2 种基金the National Natural Science Foundation of China(Grant No.61372121,61309020,61309019)the National High-Tech Research&Development Program of China(Grant No.2013AA013505)the National Science and Technology Support Program Project(Grant No.2014BAH30B01)
文摘Elastic control could balance the distributed control plane in Software-Defined Networking(SDN). Dynamic switch migration has been proposed to achieve it. However, existing schemes mainly focus on how to execute migration operation, but not why. This paper designs a decision-making mechanism based on zero-sum game theory to reelect a new controller as the master for migrated switches. It first chooses a switch for migration in the heavy controller which invites its neighbors as the game players to compete for the master role of this switch in the game-playing field(GPF) which is an occasional and loose domain for game-playing. Second, based on the concept of GPF, we design a decentralized strategy to play the game and determine which player as the final master. We implement it by extending the Open Flow protocol. Finally, numerical results demonstrate that our distributed strategy can approach elastic control plane with better performance.
基金supported by the National Natural Science Foundation of China(No.61521003)
文摘Shortest-path calculation on weighted graphs are an essential operation in computer networks. The performance of such algorithms has become a critical challenge in emerging software-defined networks(SDN),since SDN controllers need to centralizedly perform a shortest-path query for every flow,usually on large-scale network. Unfortunately,one of the challenges is that current algorithms will become incalculable as the network size increases. Therefore, inspired by the compression graph in the field of compute visualization,we propose an efficient shortest path algorithm by compressing the original big network graph into a small one, but the important graph properties used to calculate path is reserved. We implement a centralized version of our approach in SDN-enabled network,and the evaluations validate the improvement compared with the well-known algorithms.
文摘This special topic mainly focuses on the progress of physical-layer security(PLS)technologies and their potential applications for the future beyond fifth-generation(B5G)and sixth-generation(6G)networks.The tremendous growth in connectivity and the ubiquity of wireless communications have resulted in an unprecedented awareness of the importance of security and privacy.Achieving secure and trusted communications is vital for future intelligent connected applications,especially life-critical vehicle-toeverything(V2X)applications.However,the heterogeneous,dynamic and decentralized architecture of these networks leads to difficulties for cryptographic key management,and distribution.By exploiting the physical characteristics of devices,wireless channels and noise,PLS offers reliable solutions against eavesdropper attacks as complementary approaches to cryptographic techniques.
基金Project supported by the National Natural Science Foundation of China(Nos.61572520 and 61521003)。
文摘With the growing amount of information and data, object-oriented storage systems have been widely used in many applications, including the Google File System, Amazon S3, Hadoop Distributed File System, and Ceph, in which load balancing of metadata plays an important role in improving the input/output performance of the entire system. Unbalanced load on the metadata server leads to a serious bottleneck problem for system performance. However, most existing metadata load balancing strategies, which are based on subtree segmentation or hashing, lack good dynamics and adaptability. In this study, we propose a metadata dynamic load balancing(MDLB) mechanism based on reinforcement learning(RL). We learn that the Q_learning algorithm and our RL-based strategy consist of three modules, i.e., the policy selection network, load balancing network, and parameter update network. Experimental results show that the proposed MDLB algorithm can adjust the load dynamically according to the performance of the metadata servers, and that it has good adaptability in the case of sudden change of data volume.
基金the National Key Research and Development Program of China(Nos.2020YFB1806607 and 2022YFB2902202)the National Natural Science Foundation of China(Nos.61521003 and 61701538)。
文摘The sixth-generation mobile communication(6G)networks will face more complex endogenous security problems,and it is urgent to propose new universal security theories and establish new practice norms to deal with theªunknown unknownºsecurity threats in cyberspace.This paper first expounds the new paradigm of cyberspace endogenous security and introduces the vision of 6G cyberspace security.Then,it analyzes the security problems faced by the 6G core network,wireless access network,and emerging associated technologies in detail,as well as the corresponding security technology development status and the integrated development of endogenous security and traditional security.Furthermore,this paper describes the relevant security theories and technical concepts under the guidance of the new paradigm of endogenous security.
基金the National Natural Science Foundation Innovation Group Project(61521003).
文摘The common endogenous security problems in cyberspace and related attack threats have posed subversive challenges to conventional theories and methods of functional safety.In the current design of the cyber physical system(CPS),functional safety and cyber security are increasingly intertwined and inseparable,which evolve into the generalized functional safety(S&S)problem.The conventional reliability and cybersecurity technologies are unable to provide security assurance with quanti able design and veri cation metrics in response to the cyberattacks in hardware and software with common endogenous security problems,and the functional safety of CPS facilities or device has become a frightening ghost.The dynamic heterogeneity redundancy(DHR)architecture and coding channel theory(CCT)proposed by the cyberspace endogenous security paradigm could handle random failures and uncertain network attacks in an integrated manner,and its generalized robust control mechanism can solve the universal problem of quantitative design for functional safety under probability or improbability perturbation.As a generalized functional safety enabling structure,DHR opens up a new direction to solve the common endogenous security problems in the cross-disciplinary elds of cyberspace.
基金supported in part by National Key R&D Plan(2022YFB3102901)
文摘Recent advances in deep learning have led to disruptive breakthroughs in artificial intelligence(AI),fueling the jump in ChatGPT-like large language models(LLMs).As with any emerging technology,it is a two-sided coin,bringing not only vast social impacts but also significant security concerns,especially in the socio-cognitive domain.Against this back-ground,this work starts with an inherent mechanism analysis of cognitive domain games,from which it proceeds to explore the security concerns facing the cognitive domain as well as to analyze the formation mechanisms of a cognitive immune system.Finally,inspired by behavioral mimicry in biology,this work will elaborate on new approaches to cognitive security from three aspects:Mimicry Computing,Mimicry Defense,and Mimicry Intelligence.
基金supported by the National Basic Research Program of China("973"Program)(No.2013CB329104).
文摘Aiming at the problem of failure recovery in current networks,a fast failure recovery method based on equivalent cooperative routing is proposed.Firstly,the transmission path between the source and destination nodes is divided into several non-overlapping path segments.Next,backup paths are deployed for each link in the path segment through segmented routing technology,which ensures fast routing recovery after failure.Additionally,in order to avoid damaging the QoS of the data stream through the failure recovery process,the transmission is guaranteed by the intersegment QoS complement.The experimental results show that the proposed method has a low failure recovery delay under a relatively small flow table cost.
