A privacy-preserving model for multi-agent propositional planning

Abstract

Over the last years, the planning community has formalised several models and approaches to multi-agent (MA) propositional planning. One of the main motivations in MA planning is that some or all agents have private knowledge that cannot be communicated to other agents during the planning process and the plan execution. We argue that the existing models of the multi-agent planning task do not maintain the agents’ privacy when a (strict) subset of the involved agents share confidential knowledge, or when the identity/existence of at least one agent is confidential. In this paper, first we propose a model of the MA-planning tasks that preserves the privacy of the involved agents when this happens. Then we investigate an algorithm based on best first search for our model that uses some new heuristics providing a trade-off between accuracy and agents’ privacy. Finally, an experimental study compares the effectiveness of using the proposed heuristics.     

云计算环境下安全的极限学习机外包机制

应用程序中涉及到的数据日益扩大且结构日益复杂,使得在大规模数据上运行极限学习机ELM成为一个具有挑战性的任务。为了应对这一挑战,提出了一个在云计算环境下安全和实用的ELM外包机制。该机制将ELM显式地分为私有部分和公有部分,可以有效地减少训练时间,并确保算法输入与输出的安全性。私有部分主要负责随机参数的生成和一些简单的矩阵计算;公有部分外包到云计算服务器中,由云计算服务商负责ELM算法中计算量最大的计算Moore-Penrose广义逆的操作。该广义逆也作为证据以验证结果的正确性和可靠性。我们从理论上对该外包机制的安全性进行了分析。在CIFAR-10数据集上的实验结果表明,我们所提出的机制可以有效地减少用户的计算量。     

分布的缺失数据中保护隐私的贝叶斯网络学习

对隐私的保护性关注限制了参与各方对数据资源的共享使用,为此提出了从分布的缺失数据中保护隐私的贝叶斯网络学习方 法——PPHI-EM方法。该方法基于Pohlig-Hellman加密算法,使用安全有向边统计算法得到结构有向边的交集和并集。以交集作为初始网络结构,依次将并集中的其他边放入网络中,通过打分函数值的大小,判断该边是否应予保留。根据设定的适当权重,使用安全矩阵求和算法求解当前网络结构参数。循环计算直至确定网络的最优参数。该方法使用了期望统计来代替实际不存在的充分统计,使数据各方的打分函数便于分解,并基于AMS-EM方法分布迭代改进结构,使之收敛。实验结果验证了该方法的有效性。     

移动群智感知中支持隐私保护的动态激励机制

针对移动群智感知中高质量感知数据与参与用户隐私之间的矛盾,提出一种支持隐私保护的动态激励机制。首先,采用轻量级隐私保护方法,利用安全加密哈希函数为竞标用户生成不少于256位的可变地址序列,并结合随机数对候选用户节点的效用报价进行隐匿和约束;其次,通过定义区域热度、时间热度、数据完整率和数据质量等多维参数,实现任务价值与用户效用报价的动态平衡;最后,依据用户提交的效用报价和任务预算,并利用逆向拍卖思想,完成对任务参与节点的最优选择和动态激励。在群智感知系统模拟平台上进行仿真实验,结果表明所提机制不仅增强了隐私保护度和数据精确度,同时提升了时间效率和激励效果。     

Privacy preserving friend discovery cross domain scheme using re-encryption in mobile social networks

In order to guarantee the users’ privacy in the process of making friends in the mobile social networks,a new scheme of proxy re-encryption privacy protection in the cross-domain environment was introduced.The scheme employed the cross-domain multi-authority to sharing secret keys,so as to realize the access and shave of the cross-domain users data.And the secret keys of users’ attributes were re-encrypted,based on the technology of the proxy re-encryption and attribute encryption,to achieve the friends matching under the conditions of extending the access policy.Meanwhile,in purpose of enhancing the privacy of users’ data,the technology which contained the separation of users’ privacy ciphertext and secret keys was adopted.Based on that,problems in the existing system such as user data’s inability to be shared cross-cloud,less matching during the process of making friends and users’ inability to make friends when offline had been addressed.Security and experimental analysis show that this scheme can achieve chosen plaintext attack (CPA) security,ensure the privacy of friend discovery,and that is more effective than existing solutions.     

一个基于Agent的隐私保护信息过滤系统的结构

把隐私保护问题引入信息过滤系统中，提出一个改进的、基于代理的信息过滤系统的总体结构。     

On the Privacy-Preserving Outsourcing Scheme of Reversible Data Hiding over Encrypted Image Data in Cloud Computing

Advanced cloud computing technology provides cost saving and flexibility of services for users. With the explosion of multimedia data, more and more data owners would outsource their personal multimedia data on the cloud. In the meantime, some computationally expensive tasks are also undertaken by cloud servers. However, the outsourced multimedia data and its applications may reveal the data owner’s private information because the data owners lose the control of their data. Recently, this thought has aroused new research interest on privacy-preserving reversible data hiding over outsourced multimedia data. In this paper, two reversible data hiding schemes are proposed for encrypted image data in cloud computing: reversible data hiding by homomorphic encryption and reversible data hiding in encrypted domain. The former is that additional bits are extracted after decryption and the latter is that extracted before decryption. Meanwhile, a combined scheme is also designed. This paper proposes the privacy-preserving outsourcing scheme of reversible data hiding over encrypted image data in cloud computing, which not only ensures multimedia data security without relying on the trustworthiness of cloud servers, but also guarantees that reversible data hiding can be operated over encrypted images at the different stages. Theoretical analysis confirms the correctness of the proposed encryption model and justifies the security of the proposed scheme. The computation cost of the proposed scheme is acceptable and adjusts to different security levels.     

PTDC:路网环境中感知隐私的轨迹数据采集技术

针对路网环境中移动对象轨迹隐私泄露以及语义位置同质性攻击等问题,提出了一种路网环境中感知隐私的轨迹数据采集（PTDC）算法。首先,通过兴趣位置（POI）访问人次的信息墒计算路网中POI的敏感性;其次,根据顶点间敏感性和距离的混合差距,定义了θ-边权,并建立路网空间的图模型、定义了k-θ-D匿名模型以抵御语义位置同质性攻击;最后,以无向图的广度优先遍历为基础,设计了满足POI语义差异性的匿名算法,将用户的敏感采样位置用匿名区域取代,并衡量了PTDC算法处理后数据的可用性。通过实验对PTDC算法进行了验证,并和自由空间中的基于语义位置的隐私保护算法——YCWA进行了比对。理论上讲,YCWA算法的隐私保护度低于PTDC算法。实验表明,PTDC算法的信息丢失率平均在15%左右,空间范围查询误差平均在12%左右,略逊于YCWA算法;然而,PTDC算法的运行时间在5 s以内,远远优于YCWA算法,可满足实时在线数据采集的需求。     

面向企业信用评估的分布式隐私保护数据挖掘

介绍企业信用评估和当前隐私保护数据挖掘技术的最新进展,利用适用于企业信用评估的大规模分布式隐私保护数据挖掘架构,讨论了基于该架构的面向企业信用评估的分布式隐私保护数据挖掘。该研究不仅将有助于大规模分布式环境下的隐私保护数据挖掘系统的研发,而且能够有力推动“信用中国”的建设步伐,以达到更好地服务经济的目的。     

联邦学习中的隐私保护技术

联邦学习是顺应大数据时代和人工智能技术发展而兴起的一种协调多个参与方共同训练模型的机制.它允许各个参与方将数据保留在本地,在打破数据孤岛的同时保证参与方对数据的控制权.然而联邦学习引入了大量参数交换过程,不仅和集中式训练一样受到模型使用者的威胁,还可能受到来自不可信的参与设备的攻击,因此亟需更强的隐私手段保护各方持有的...