针对传统的网络拓扑构建方法难以很好描述推荐平台中网络用户的情况,提出了一种基于用户"围观"行为的网络构建方法,对构建的网络进行分析.由于传统的标签传播算法易发生标签入侵和算法结果不稳定,改进标签传播算法的标签更新规则,提出一种适用于挖掘推荐网络社区的标签传播算法.利用短视频推荐平台数据集,对所提算法进行了实验论证.结果表明,所提算法挖掘的社区内部兴趣相似度显著高于整个网络中的兴趣相似度,有较高的兴趣内聚性.

In order to solve the key security problem in attribute-based encryption, a ciphertext policy attribute-based encryption scheme against key abuse was proposed based on the ring learning with error over ring and the access structure of ordered binary decision diagram.Firstly, two different institutions were constructed to jointly generate the user’s secret key, which reduced the risk of key disclosure by institutions.Secondly, the user’s specific information was embedded in each secret key to realize the traceability of the key, and the access of illegal users and malicious users were avoided by maintaining the white list.In addition, the access structure of ordered binary decision diagram was adopted by the proposed scheme, and the positive and negative values of attributes on the basis of supporting attribute AND, OR and Threshold operation were increased.Analysis shows that the proposed scheme meets the distinguishable security of anti-collusion attack and chosen-plaintext attack, reduces the storage and computing overhead, and it is more practical than other schemes.

Aiming at the problem of illegal data sharing of malicious users in the access control scheme based on attribute-based encryption, an access control scheme that can restrict the sending ability of data owners is proposed. By adding a sanitizer to sanitize the ciphertext, it can ensure that parties who do not adhere to the system control policy cannot share information effectively. The scheme is constructed based on blockchain, and the traceability of access process can be realized. Off-chain storage can also lower the blockchain storage load. The scheme meets the No-Read and No-Write rules, achieves chosen-plaintext attack security under the random oracle model, and can against quantum attacks. As a result of theoretical analysis and experimental simulation, the scheme has certain feasibility and practical significance.

In view of the fact that the traditional information dissemination model based on network topology is difficult to describe the information dissemination in the recommendation platform well, by investigating and studying the information dissemination characteristics of the recommendation platform, an information dissemination model of the recommendation platform based on the network community is designed. The simulation results of the model are compared with the real information dissemination of the recommendation platform. With the support of data, the Pearson similarity between the simulation results and the real information dissemination curve reaches 0.99. Finally, the influence of information reception rate, acceptance rate, recommended community density, and community dissemination threshold on information dissemination is explored through simulation experiments, which provides a certain theoretical support for the management and control of public opinion information in the recommendation platform.

随着信息传播方式的改变,通过信息推荐平台的谣言信息传播成为了谣言传播的重要方式,构建推荐模式下的谣言传播模型对网络谣言的治理具有积极作用.考虑到推荐机制对用户的分割效应,在谣言传播过程中根据人物相似性将网络用户划分在不同的传播域,定义了传播域中用户与信息的交互方式,提出了一种基于谣言信息热度和平台用户密度的谣言传播模型,并对影响谣言信息传播的因素进行了仿真分析.通过仿真发现:在推荐机制下,人们对信息的接受率主要是受信息本身影响;在推荐平台上,增大谣言信息传播阈值有利于阻止谣言的传播;谣言的接受率越大,对辟谣信息的加入时间要求越高.

Aiming at the shortcomings of traditional label propagation algorithms that are prone to label invasion and unstable results, the label updating rules of label propagation algorithms are improved, and a label propagation algorithm suitable for mining dense network communities is proposed. Finally, using the short video recommendation platform data set, the algorithm proposed in this paper is experimentally demonstrated in terms of interest similarity, modularity and division tightness indicators. The results show that the similarity of interest within the community mined by the proposed algorithm is significantly higher than the similarity of interest in the entire network, and has a higher interest cohesion.

Information hiding technology is a technology that transmits secret information through a carrier, and is a research hotspot in the field of information security. The traditional information hiding algorithm relies on the meticulous design of human beings, and obtains the dense image through modification. With the development of deep learning, the integration of information hiding technology and deep learning has resulted in many information hiding technologies based on deep learning. Among them, image hiding has become a research hotspot due to its large steganographic capacity. Therefore, this paper reviews the information hiding technology based on deep learning. According to the difference of hidden models, it is analysed from four aspects: (1) information hiding model based on encoder-decoder; (2) information hiding model based on generative adversarial network; (3) information hiding model based on invertible network; (4) information hiding model based on neural network information hiding models for style transfer. Finally, these models are analysed and compared, and the future development direction is discussed and prospected.

The network security situation is grim, and the problem of "information isolated island" is becoming increasingly prominent. In view of the low efficiency and insufficient security of data cross-domain sharing in the open network environment, a searchable data sharing scheme supporting cross-domain is proposed based on attribute encryption technology. Firstly, different types of nodes on the blockchain are used to realize the data sharing of users in different domains. Secondly, the flexible ciphertext-search function is realized through the search form of keyword strategy. Moreover, the scheme adopts the mode of storage under the chain, which reduces the operation pressure of the blockchain. At the same time, according to the characteristics of the blockchain, the traceability and tamper-proof of the access process can be realized. Finally, the analysis shows that the scheme can resist quantum attack and collusive attack while avoiding complex bilinear operation and meet the security of trapdoor search and indistinguishability under chosen-plaintext attack. Compared with other searchable attribute-based encryption schemes, the scheme has certain advantages in function and performance.

针对智能终端在军事领域应用中存在的数据安全问题,提出了一种战术环境下基于边缘计算的访问控制方案,通过采用属性加密和国密SM4算法相结合的混合加密模式,在对数据加密保护的同时不需要依赖第三方即可实现对用户的访问控制,且方案在解密过程中可依靠节点辅助解密,减少了用户解密的开销.最后对方案进行了理论分析和仿真实验,结果表明方案具有一定的安全性和可行性.

目前大多数的轨迹隐私保护方法对轨迹的形状相似性考虑并不充分,并且容易忽略各轨迹点之间的时序相关性,导致生成的干扰轨迹可用性不高.为了解决这些问题,提出了一种基于密度的噪声应用空间聚类(density based spatial clustering of application with noise,DBSCAN)算法的差分隐私轨迹保护机制.首先,使用DBSCAN算法对数据进行聚类分析,降低数据集中噪声点对聚类效果的影响;其次,根据用户活动轨迹点的时序关系,生成位置转移概率矩阵,利用差分隐私的方法确保生成的干扰轨迹点与真实轨迹点具有相似的位置转移概率;最后,综合考虑差分隐私预算和弗朗明歇距离(Fréchet distance)对轨迹相似性的影响,选取位置干扰点.通过仿真实验分析,方案在效率上具有明显的优势,并且生成的干扰轨迹与真实的位置轨迹相比具有较高的形状相似性.

In order to solve the problem of key abuse in attribute-based encryption, a traceable ciphertext-policy attribute-based encryption scheme is proposed based on the learning with error over ring and the ordered binary decision diagram structure. Firstly, the specific information of the user is embedded in each private key to realize the traceability of the key. Secondly, the access of illegal users and malicious users is avoided by maintaining the white list. In addition, the scheme adopts the access structure of ordered binary decision diagram, and increases the positive and negative values of attributes on the basis of supporting the operation of “AND”, “OR” and “Threshold”. Finally, through analysis, the scheme meets the security of anti-collusion attack and chosen-plaintext attack security, reduces the storage and computing overhead, and is more practical than other schemes.

As one of the ways to reflect the views of the masses in modern society, online reviews have great value in public opinion research. The analysis of potential public opinion information from online reviews has a certain value for the government to clarify the next work direction. In this paper, the event evolution graph is designed to make COVID-19 network public opinion prediction. The causal relationship was extracted in the network reviews after the COVID-19 incident to build an event evolution graph of COVID-19 and predict the possibility of the occurrence of the derivative public opinion. The research results show the hot events and evolution direction of COVID-19 network public opinion in a clear way, and it can provide reference for the network regulatory department to implement intervention.

针对属性加密方案中的运行效率和属性更新问题,提出了一种基于环上误差学习问题(Learning With Error over Ring,RLWE)的可撤销分层属性加密方案.方案通过多等级的门限秘密共享矩阵将属性进行分层,权限等级高的属性恢复秘密的能力大于权限等级低的属性,且高权限等级不可被替代;另外,方案实现了属性级的用户撤销,基于第三方机构通过控制用户对属性陷门的获取降低了系统的计算开销.该方案能抵抗用户合谋攻击且满足随机预言机模型下的选择明文安全,与现有方案对比,在实现了属性分层的同时增加了属性撤销的功能,并在多项式环上进行运算,提高了加解密效率,对实际应用场景有更好的适应性.