基于云原生数据库的许多应用场景需要处理海量的数据流.为了实时分析数据流中的群体趋势信息而又不泄露单个用户的隐私,这些应用需要在每个时刻都可以为数据流中的最近数据集快速创建可以安全发布的差分隐私直方图.然而,现有的直方图发...基于云原生数据库的许多应用场景需要处理海量的数据流.为了实时分析数据流中的群体趋势信息而又不泄露单个用户的隐私,这些应用需要在每个时刻都可以为数据流中的最近数据集快速创建可以安全发布的差分隐私直方图.然而,现有的直方图发布方法因缺乏高效数据结构,导致无法快速提取关键信息以确保数据的实时可用性.为解决此问题,深入分析数据采样与隐私保护之间的关系,提出基于采样的数据流差分隐私快速发布算法SPF(sampling based fast publishing algorithm with differential privacy for data stream).SPF首创高效数据流采样草图结构(efficient data stream sampling sketch structure,EDS),EDS对滑动窗口内数据进行采样统计估计,并过滤不合理数据,实现了对关键信息的快速提取.然后,证明EDS结构输出的近似值理论上等效于对真实值添加差分隐私噪声.最后,为了满足用户所提供的隐私保护强度,并且避免正确反映原始数据流的真实情况,提出了一种基于高效数据流采样的自适应加噪算法.根据用户的隐私保护强度和EDS结构所提供的隐私保护强度之间的关系,通过隐私分配的方式自适应生成最终可发布直方图.实验证明,相较于现有算法,SPF在保持相同数据可用性的前提下显著降低了时间和空间开销.展开更多
针对现实不确定数据流具备分布非凸性和包含大量噪声等特点,提出不确定数据流聚类算法Clu_Ustream(clustering on uncertain stream)来解决对近期数据进行实时高效聚类演化问题。首先,在线部分利用子窗口采样机制采集滑动窗口中的不确...针对现实不确定数据流具备分布非凸性和包含大量噪声等特点,提出不确定数据流聚类算法Clu_Ustream(clustering on uncertain stream)来解决对近期数据进行实时高效聚类演化问题。首先,在线部分利用子窗口采样机制采集滑动窗口中的不确定流数据,采用双层概要统计结构链表存储概率密度网格的统计信息;然后,离线聚类过程中通过衰减窗口机制弱化老旧数据的影响,并定期对窗口中的过期子窗口进行清理;同时采用动态异常网格删除机制有效过滤离群点,从而降低算法的时空复杂度。在模拟数据集和网络入侵真实数据集上的仿真结果表明,Clu_Ustream算法与其他同类算法相比具有较高的聚类质量和效率。展开更多
The special purpose Monte Carlo program McMesh was used to study neutron transport in coal slurries for on stream determination of the slurry parameters. McMesh uses the mesh weight window method as the major variance...The special purpose Monte Carlo program McMesh was used to study neutron transport in coal slurries for on stream determination of the slurry parameters. McMesh uses the mesh weight window method as the major variance reduction technique with other methods such as exponential transforms and correlated sampling included as options. There was good agreement between the calculated results from McMesh and from MCNP, a general Monte Carlo program, but McMesh was more efficient and more convenient.展开更多
Flying plots detection has been the focus of relay protection in power system for a long time. With the promotion of Smart substation in our country, the number of SV devices is greatly increased. Abnormal data (flyin...Flying plots detection has been the focus of relay protection in power system for a long time. With the promotion of Smart substation in our country, the number of SV devices is greatly increased. Abnormal data (flying plot) caused by sampling device itself has brought tremendous pressure to the power system. The traditional flying plot detection algorithm has plenty of defects, such as low pertinence, low sensitivity and long sampling period. This paper proposes a new algorithm to identify flying plot by analyzing the wave form characteristics of sampling data. The traditional waveform recognition methods are combined in this algorithm. It has the concept of standard wave window and can distinguish flying plot in a short time. In addition, sine recovery algorithm is used to recover the flying plot. This paper uses PSCAD software to verify the validity of this algorithm. Simulation results show that the proposed method has high reliability.展开更多
文摘基于云原生数据库的许多应用场景需要处理海量的数据流.为了实时分析数据流中的群体趋势信息而又不泄露单个用户的隐私,这些应用需要在每个时刻都可以为数据流中的最近数据集快速创建可以安全发布的差分隐私直方图.然而,现有的直方图发布方法因缺乏高效数据结构,导致无法快速提取关键信息以确保数据的实时可用性.为解决此问题,深入分析数据采样与隐私保护之间的关系,提出基于采样的数据流差分隐私快速发布算法SPF(sampling based fast publishing algorithm with differential privacy for data stream).SPF首创高效数据流采样草图结构(efficient data stream sampling sketch structure,EDS),EDS对滑动窗口内数据进行采样统计估计,并过滤不合理数据,实现了对关键信息的快速提取.然后,证明EDS结构输出的近似值理论上等效于对真实值添加差分隐私噪声.最后,为了满足用户所提供的隐私保护强度,并且避免正确反映原始数据流的真实情况,提出了一种基于高效数据流采样的自适应加噪算法.根据用户的隐私保护强度和EDS结构所提供的隐私保护强度之间的关系,通过隐私分配的方式自适应生成最终可发布直方图.实验证明,相较于现有算法,SPF在保持相同数据可用性的前提下显著降低了时间和空间开销.
文摘针对现实不确定数据流具备分布非凸性和包含大量噪声等特点,提出不确定数据流聚类算法Clu_Ustream(clustering on uncertain stream)来解决对近期数据进行实时高效聚类演化问题。首先,在线部分利用子窗口采样机制采集滑动窗口中的不确定流数据,采用双层概要统计结构链表存储概率密度网格的统计信息;然后,离线聚类过程中通过衰减窗口机制弱化老旧数据的影响,并定期对窗口中的过期子窗口进行清理;同时采用动态异常网格删除机制有效过滤离群点,从而降低算法的时空复杂度。在模拟数据集和网络入侵真实数据集上的仿真结果表明,Clu_Ustream算法与其他同类算法相比具有较高的聚类质量和效率。
文摘The special purpose Monte Carlo program McMesh was used to study neutron transport in coal slurries for on stream determination of the slurry parameters. McMesh uses the mesh weight window method as the major variance reduction technique with other methods such as exponential transforms and correlated sampling included as options. There was good agreement between the calculated results from McMesh and from MCNP, a general Monte Carlo program, but McMesh was more efficient and more convenient.
文摘Flying plots detection has been the focus of relay protection in power system for a long time. With the promotion of Smart substation in our country, the number of SV devices is greatly increased. Abnormal data (flying plot) caused by sampling device itself has brought tremendous pressure to the power system. The traditional flying plot detection algorithm has plenty of defects, such as low pertinence, low sensitivity and long sampling period. This paper proposes a new algorithm to identify flying plot by analyzing the wave form characteristics of sampling data. The traditional waveform recognition methods are combined in this algorithm. It has the concept of standard wave window and can distinguish flying plot in a short time. In addition, sine recovery algorithm is used to recover the flying plot. This paper uses PSCAD software to verify the validity of this algorithm. Simulation results show that the proposed method has high reliability.
