The pervasive adoption of the Internet of Things (IoT) is accompanied by numerous network security threats, making the timely detection of anomalies in traffic data through intrusion detection increasingly critical. The existing intrusion detection methods based on federated learning can achieve good results under the condition of sufficient labeled data. However, the traffic data of participants in real IoT environments have various characteristics and there are a large amount of unlabeled data, which easily leads to the performance degradation of the intrusion detection model. To address this challenge, this paper proposes a novel federated transfer learning approach for IoT intrusion detection based on multi-party collaboration called MPdetector. MPdetector uses an encoder to extract the feature representation of diverse traffic data from heterogeneous clients, and maps the feature representation of each client to a unified feature space. In addition, a label transfer strategy is introduced to make full use of unlabeled data, and a new mapping function is used to reconstruct the traffic data of each client to expand client's local data set, which can further improve the detection performance of the model in varied and complex IoT environments. Theoretical analysis proves that the entire transfer learning process of MPdetector is conducted within a secure context. Experiments on four widely used intrusion detection datasets show that MPdetector can detect known and unknown abnormal traffic more accurately than the existing three classical intrusion detection algorithms, and has strong generalization. Meanwhile, the detection effect of MPdetector will be further improved with the increase of the volume of labeled traffic.

Federated learning in heterogeneous environments faces the fairness problem of global models performing differently across participants due to differences in participant sample resources and participation. The existing fairness algorithms which mainly adjust the aggregate weight are easy to give malicious participants more attack space, resulting in reduced robustness of the model. The mainstream attack defense schemes are mostly based on the assumption of participants’ data with independence and same distribution, which easily filters out legal but uncommon updates, leading to more unfairness. Aiming at this contradiction, this article proposes a fair and robust federated learning approach called FLClustering. Based on similar model parameters, a cluster of participants is generated to mitigate the heterogeneity among participants in the same cluster. Then a Byzantine attack defense mechanism is introduced in the cluster, which can resist label flipping attacks, model poisoning attacks and other Byzantine attacks while avoiding the error of discarding legitimate participants. Especially, both a dynamic adjustment strategy and a fair participant selection mechanism are proposed to further reduce the performance difference between participants. Experiments on Modified National Institute of Standards and Technology database (MNIST) and Fashion-MNIST datasets show that the proposed approach can obtain a fair and robust global model. Compared with the existing fairness algorithms, the accuracy under attack is improved by more than 3% and the standard deviation is reduced by more than 4%.

Cloud-edge collaborative federated learning allows user terminals, edge servers and the cloud to jointly train a model without exchanging raw data, thus preserving privacy. However, traditional schemes weight updates only by dataset size, ignoring edge-training accuracy and inference-cost sensitivity, which leads to high classification expense and low edge accuracy. Meanwhile, tampered updates and highly skewed sample sizes further degrade performance. To address these issues, this paper proposes an Efficient cloud-edge federated-learning framework with bidirectional integrity protection called EBIPFL, where Merkle trees are employed to verify the integrity of encrypted model updates in both uplink and downlink directions. A lightweight small-sample partitioning strategy is presented to reduce on-device workload, while the aggregation rule fuses cost-sensitive error, precision, recall and trainer count to produce a more informed global model. Finally, theoretical analysis and experimental evaluations are conducted. The results demonstrate that EBIPFL not only improves model accuracy while maintaining acceptable storage and time costs, but also effectively defends against tampered update attacks.

Machine Learning as a Service (MLaaS) has been widely used, such as logistic regression models trained in the cloud to provide people with classification predictions. However, these services allow the cloud to collect data from a large number of users, which could raise privacy concerns. With sharing of local models rather than user data, Federated Learning (FL) alleviates data privacy. But, in FL setting the difference between the predictive value provided by the optimal model and the shared global model (called maturing model) that approximates it is not significant, which still raises a security risk. In this paper, we design a flexible and privacy-preserving FL system to tackle these issues. This system makes the training process divide into two stages by analyzing whether the metrics of the iterative global gradient reach a set threshold or not. The two stages are implemented with Re-Encryption and CKKS homomorphism, which not only enables a secure aggregation in the cloud, but also maintains the security of the maturing model. Through detailed analysis, we prove the security of our protocol. Moreover, quantitative comparisons in the final experiments demonstrates that our scheme can achieve trade-off between accuracy, training time cost and communication overhead.

Federated learning has the problem of privacy leakage from the gradient. The existing gradient protection schemes based on homomorphic encryption incur a large time cost and the risk of gradient leakage caused by potential collusion between participants and aggregation server. A new federated learning method called FastProtector is proposed, where the idea of SignSGD is introduced when homomorphic encryption is used to protect participant gradients. Exploiting the feature that the majority of positive and negative gradients determine the aggregation result to make the model convergent, the gradient is quantified and the gradient updating mechanism is improved, which can reduce the overhead of gradient encryption. Meanwhile,an additive secret sharing scheme is proposed to protect the gradient ciphertext against collusion attacks between malicious aggregation servers and participants. Experiments on MNIST and CIFAR-10 dataset show that the proposed method can reduce the total encryption and decryption time by about 80% while ensuring high model accuracy.

Federated Learning is a distributed machine learning framework, which mainly adopts cloud-edge collaborative computing mode and supports multiple participants to train models without directly sharing local data. However, participants’ sensitive information may still be leaked through their gradients. Besides, incorrect aggregated results returned by the aggregation server may reduce the effect of joint modeling. This article proposes a privacy-preserving and verifiable federated learning method called PPVerifier to support privacy protection and verification of aggregated results in the cloud-edge collaborative computing environment. By the integrating Paillier homomorphic encryption and random number generation technique, all gradients and their ciphertexts can be protected. Meanwhile, an additive secret-sharing scheme is introduced to resist potential collusion attacks among the aggregation server, malicious participants, and edge nodes. Moreover, a verification scheme based on discrete logarithm is proposed, which can not only verify the correctness of aggregated results, but also discover lazy aggregation servers, and the verification overhead can be reduced by over half compared with a bilinear aggregate signature method. Finally, theoretical analysis and experiments conducted on the MNIST data set prove that our proposed method can achieve gradient protection and correctness verification of the aggregated results with higher efficiency.

针对大数据应用中用户共享数据的访问控制由半可信云服务商实施所带来的隐私泄露、策略和访问日志易被篡改等问题,提出一种基于区块链的策略隐藏大数据访问控制方法(A policy-hidden big data access control method based on blockchain,PHAC).该方法采用区块链技术实施访问控制以减少对服务商的信任依赖,引入属性基加密(Attribute-based encryption,ABE)以及双线性映射技术,实现在不泄露访问控制策略的前提下,通过智能合约正确执行访问控制策略.同时,解耦访问控制策略,简化用户策略的发布、更新和执行.并应用链上和链下存储相结合方式,解决智能合约和访问控制策略占用区块链节点资源不断增大的问题.最后,对该方法进行了理论分析和HyperLedger Fabric环境下的实验评估,结果表明该方法能在策略隐藏情况下有效实现访问控制,但不会给数据拥有者、区块链节点增加过多额外计算和存储开销.

Edge-assisted Internet of things applications often need to use cloud virtual network services to transmit data. However, the internal threats such as illegal management and configuration to cloud platform intentionally or unintentionally will lead to virtual network security problems such as malicious changes of user network and hijacked data flow. It will eventually affect edge-assisted Internet of things applications. We propose a virtual network internal threat detection method called VNGuarder in a cloud computing environment, which can effectively monitor whether the virtual network configuration of legitimate users under the IaaS cloud platform has been maliciously changed or destroyed by insiders. First, based on the life cycle of cloud virtual network services, we summarized two types of internal attacks involving illegal use of virtualization management tools and illegal invocation of virtual network-related processes. Second, based on normal behavior of tenants, a hierarchical trusted call correlation scheme is proposed to provide a basis for discovering that insiders illegally call virtualized management tools and virtual network-related processes on the controller node of the cloud platform or the network node and compute node. Third, a trace-enable mechanism combining real-time monitoring and log analysis is introduced. By collecting and recording the complete call process of virtual network management and configuration in the cloud platform, and comparing it with the result of the hierarchical trusted call correlation, abnormal operations can be reported to the tenants in time. Comprehensive simulation experiments on the Openstack platform show that VNGuarder can effectively detect illegal management and configuration of virtual networks by insiders without significantly affecting the creation time of tenant networks and the utilization of CPU and memory.

How to audit the integrity of data stored on the cloud with incomplete trust is an important problem that restricts the development of cloud storage. Although there are several data integrity audit schemes in cloud storage, the increased need to protect sensitive information and support large-scale data storage and dynamic update will result in a significant increase in audit cost, which seriously affects the efficiency of existing cloud audit systems. To solve this problem, we propose Ldasip, a lightweight dynamic auditing method that supports sensitive information protection in cloud storage. Exploiting identity-based data integrity audit, a data masking technology is introduced into to protect user’s sensitive information. At the same time, an improved multibranch tree structure is proposed to realize dynamic audit and reduce communication overhead in the verification process. Theoretical analysis and comprehensive experiments have been conducted, which demonstrate the effectiveness of Ldasip. The results show that Ldasip can ensure the correctness of the audit, protect the sensitive information in the user's stored content, and support the dynamic update of data with less audit time and communication overhead.

How to detect malicious insiders' improper access to tenant data has become more crucial in IaaS cloud environment, especially with the cloud administrators gaining more control on customers' virtual machines and data in reality. In this paper, we propose an insider threats detection approach based on behavior traceability called BTDetect. First, we analyze the service invocation interfaces of IaaS cloud environment, such as computing service, remote call, management implementation and virtualization management, and condense the complete process of cloud user behavior. Using tree-based modeling technique, a behavior-tree construction algorithm is proposed to construct the normal behavior tree that can describe various legal operations of cloud users. Second , we set up trace points of cloud service behavior on multi-layer cloud service APIs, then we collect information of each interface being invoked across multiple nodes. Third, we use underlying virtualization behavior keyword matching technology to match the collected behaviors with the user's normal behavior tree and then the malicious internal threat can be identified through tree-based integrity analysis. Finally, some experiments are conducted to evaluate the feasibility and veracity of the proposed method in Openstack platform. The results suggest that our method can not only identify internal threat but also have high recognition rate.

针对云环境下分布式拒绝服务(distributed denial-of-service,DDoS)攻击加密攻击流量隐蔽性更强、更容易发起、规模更大的问题,提出了一种云环境下基于信任的加密流量DDoS发现方法TruCTCloud.该方法在现有基于机器学习的DDoS攻击检测中引入信任的思想,结合云服务自身的安全认证,融入基于签名和环境因素的信任评估机制过滤合法租户的显然非攻击流量,在无需对加密流量解密的前提下保障合法租户流量中包含的敏感信息.其后,对于其他加密流量和非加密流量,引入流包数中位值、流字节数中位值、对流比、端口增速、源IP增速这5种特征,基于特征构建Ball-tree并提出基于k近邻(k-nearest neighbors,kNN)的流量分类算法.最后,在OpenStack云环境下检测了提出方法的效果,实验表明TruCTCloud方法能快速发现异常流量和识别DDoS攻击的早期流量,同时,能够有效保护合法用户的敏感流量信息.

Since the widespread adoption of edge computing and IoT technology, Control-Flow Hijacking (CFH) attacks targeting programs in resource-constrained embedded devices have become prevalent. While the Coarse-Grained Control-Flow integrity Attestation (CGCFA) lacks accuracy for the CFH attacks detection, the Fine-Grained Control-Flow integrity Attestation (FGCFA) detect the attacks more accurately but with high overheads, which can be a big burden (e.g., to industrial control system with strict performance requirements). In this paper, we propose a NSGA-II (Nondominated Sorting Genetic Algorithm-II) based Granularity-Adaptive Control-Flow Attestation (GACFA) for the programs in embedded devices. Specifically, we propose a Granularity-Adaptive Control-Flow representation model to reduce the complexity of programs’ control-flow graph and propose NSGA-II-based granularity-adaptive strategy generation algorithm to balance the security and performance requirements. Besides, runtime protection for the GACFA at the program end with SGX is proposed to protect the integrity and confidentiality of control-flow measurement data. The experiments show that our work can find out the best-so-far control-flow granularity with stability and provide secure program attestation for the verifier. In addition, the security/performance benefit of adopting our proposal over CGCFA is 13.7, 25.1, and 43.0 times that of adopting FGCFA over ours in different threat scenarios.

Since the widespread adoption of edge computing and IoT technology, Control-Flow Hijacking (CFH) attacks targeting programs in resource-constrained embedded devices have become prevalent. While the Coarse-Grained Control-Flow integrity Attestation (CGCFA) lacks accuracy for the CFH attacks detection, the Fine-Grained Control-Flow integrity Attestation (FGCFA) detect the attacks more accurately but with high overheads, which can be a big burden (e.g., to industrial control system with strict performance requirements). In this paper, we propose a NSGA-II (Nondominated Sorting Genetic Algorithm-II) based Granularity-Adaptive Control-Flow Attestation (GACFA) for the programs in embedded devices. Specifically, we propose a Granularity-Adaptive Control-Flow representation model to reduce the complexity of programs’ control-flow graph and propose NSGA-II-based granularity-adaptive strategy generation algorithm to balance the security and performance requirements. Besides, runtime protection for the GACFA at the program end with SGX is proposed to protect the integrity and confidentiality of control-flow measurement data. The experiments show that our work can find out the best-so-far control-flow granularity with stability and provide secure program attestation for the verifier. In addition, the security/performance benefit of adopting our proposal over CGCFA is 13.7, 25.1, and 43.0 times that of adopting FGCFA over ours in different threat scenarios.

访问控制是网络空间安全的一种共性技术,访问控制理论是网络空间安全学科所特有的理论基础.为了使访问控制课程教学跟上国内外相关学科的发展趋势,学生能够对前沿学术动态有所了解并参与国际交流,实行双语教学是必然的选择.结合近两年硕士研究生"访问控制理论与实践"双语教学改革的探索实践,对教学参考资料选用、课程大纲和教学内容优化以及多元化教学方法等课堂实践经验进行总结,并对双语教学中遇到的问题提出若干思考和建议.

数据外包云存储是当前主流的海量数据存储方式,这种模式下用户失去对其数据的绝对控制权,恶意租户或云服务内部人员可能会篡改或破坏云端数据,如何确保非可信云存储环境下外包数据的完整性是制约云计算发展的重要安全问题.文章对现有的数据完整性审计工作进行综述,基于对外包数据完整性审计问题的统一抽象给出审计模型分类,结合支撑数据动态操作、计算和通信开销等审计目标,从审计中用户关键数据是否共享的视角分类,对当前典型方案进行了对比分析和优缺点讨论,最后总结了外包数据完整性审计的未来发展方向.

Data privacy protection is crucial to cloud computing since privacy leakage may prevent users from using cloud services. To ensure data privacy, we propose PriGuarder, a novel privacy-aware access control method. This method spans the three stages of a cloud service, i.e., user registration, data creation, and data access. At each stage, users can choose two modes to interact with the cloud service provider, i.e., direct or indirect. With the indirect mode, an attribute fuzzy grouping scheme is introduced to ensure user identity privacy and attribute privacy in all the three stages. Furthermore, exploiting data encryption and timestamp techniques, new access control protocols are proposed to regulate interactions between users and the cloud service provider. We illustrate the use of our method in the context of Amazon S3. Theoretical analysis and comprehensive simulation experiments have been conducted, which demonstrate the efficacy of PriGuarder.

With the rapid development of android smart terminals, android applications are exhibiting explosive growth.However, there remains a challenging issue facing android system, a malicious application may broadcast user's private information.In this paper, we propose a Naive Bayesian-based approach for analyzing private information leakage under the android broadcast mechanism, which calls BRbysA.Firstly, broadcast actions registered in manifest.xmlare picked up statically by keyword matching technique.Secondly, with the Xposed framework, the broadcast actions specified at run time are discovered by hooking broadcast callback onReceive() function.Combining the above two ways, we can capture all real-time broadcast actions in an android application.Thirdly, we adopt Naive Bayesian learning algorithm, all broadcast actions which involved in users' privacy leakage are analyzed and classified.Finally, we evaluate the proposed approach by using the dataset from Drebin and Google Play.

It is very important for crowdsourcing system to analyze crowdsourcing workers' collaborative behaviors.In this paper,the evolutionary process of crowdsourcing quality was studied based on classic hierarchically-organized mode.By establishing an evolutionary game model for crowdsourcing task collaboration among different virtual organizations,the evolutionary stability of crowdsourcing systems was analyzed and the dynamics of crowdsourcing workers' behaviors were discussed macroscopically.The key factors affecting the evolution of crowdsourcing system,including the economic benefits,consumer utilities from completing crowdsourcing tasks and risks from insecure participating,were suggested and how the factors work was presented.All these results together provide the theory basis for designing quality control methods of crowdsourcing system.

针对信息安全体系结构课程的教学目标和现状特点,基于混合式教学理念并结合最新的慕课教学模式,研究慕课教学与信息安全体系结构课程课堂教学之间的关系,设计一种包含创建并搜集慕课网络资源、开发课程内容与教学模块、教学过程的具体实施到课程测评反馈四个教学环节的混合式教学模式.该教学模式将课上与课下、线上与线下紧密融合,在发挥教师教学主导作用的同时,突出了学生学习的主体性地位,是提高信息安全体系结构课程教学实效性的有效模式.

In the recent fifth generation (5G) research, filter bank multi-carrier (FBMC) technique is one of the competitive substitutes to orthogonal frequency division multiplexing (OFDM). The conventional FBMC schemes, such as FBMC staggered-modulated multitone (FBMC-SMT) and frequency spreading FBMC (FS-FBMC), have the intrinsic time-domain interference which is caused by the overlap of symbols and depend on offset quadrature amplitude modulation (OQAM). In order to overcome the two drawbacks, a non-overlapping QAM-FBMC scheme is proposed based on the FBMC-SMT and analyzed in this paper. Moreover, the non-overlapping QAM-FBMC scheme not only discards cyclic prefix (CP) as other FBMC schemes, but also shortens the duration length of symbol waveforms compared with that of FS-FBMC. Therefore, its spectrum efficiency (SE) outperforms other multi-carrier systems. Simulation results indicate that non-overlapping QAM-FBMC scheme indeed improves the SE and obtains almost the same bit error rate performance as that of the FBMC-SMT as well as OFDM in multi-path fading channel.