目前针对未知的Android恶意应用可以采用机器学习算法进行检测,但传统的机器学习算法具有少于三层的计算单元,无法充分挖掘Android应用程序特征深层次的表达。文中首次提出了一种基于深度学习的算法DDBN(Data-flow Deep Belief Network)...目前针对未知的Android恶意应用可以采用机器学习算法进行检测,但传统的机器学习算法具有少于三层的计算单元,无法充分挖掘Android应用程序特征深层次的表达。文中首次提出了一种基于深度学习的算法DDBN(Data-flow Deep Belief Network)对Android应用程序数据流特征进行分析,从而检测Android未知恶意应用。首先,使用分析工具Flow Droid和SUSI提取能够反映Android应用恶意行为的静态数据流特征;然后,针对该特征设计了数据流深度学习算法DDBN,该算法通过构建深层的模型结构,并进行逐层特征变换,将数据流在原空间的特征表示变换到新的特征空间,从而使分类更加准确;最后,基于DDBN实现了Android恶意应用检测工具Flowdect,并对现实中的大量安全应用和恶意应用进行检测。实验结果表明,Flowdect能够充分学习Android应用程序的数据流特征,用于检测未知的Android恶意应用。通过与其他基于传统机器学习算法的检测方案对比, DDBN算法具有更优的检测效果。展开更多
流量分类是优化网络服务质量的基础与关键.机器学习算法利用数据流统计特征分类流量,对于识别加密私有协议流量具有重要意义.然而,特征偏置和类别不平衡是基于机器学习的流量分类研究所面临的两大挑战.特征偏置是指一些数据流统计特征...流量分类是优化网络服务质量的基础与关键.机器学习算法利用数据流统计特征分类流量,对于识别加密私有协议流量具有重要意义.然而,特征偏置和类别不平衡是基于机器学习的流量分类研究所面临的两大挑战.特征偏置是指一些数据流统计特征在提高部分应用识别准确率的同时也降低了另外一部分应用识别的准确率.类别不平衡是指机器学习流量分类器对样本数较少的应用识别的准确率较低.为解决上述问题,提出了基于集成聚类的流量分类架构(traffic classification framework based on ensemble clustering,简称TCFEC).TCFEC由多个基于不同特征子空间聚类的基分类器和一个最优决策部件构成,能够提高流量分类的准确率.具体而言,与传统的机器学习流量分类器相比,TCFEC的平均流准确率最高提升5%,字节准确率最高提升6%.展开更多
In this paper we summarize the characteristics of the dishpan experiment, the principle of substance revolving, and the scientific basis of the “retrograde wave in only one direction” with respect to weather data an...In this paper we summarize the characteristics of the dishpan experiment, the principle of substance revolving, and the scientific basis of the “retrograde wave in only one direction” with respect to weather data and S. C. OuYang＇s articles in which the fundamental questions in the meteorological theory were pointed out. Furthermore, we discuss the systematic changes involving the concept, theory, and method that substance evolves.展开更多
Based on the observation data and the reanalysis datasets, the variability and the circulation features influencing precipitation in the Tibetan Plateau (TP) are investigated. Taking into account the effects of topogr...Based on the observation data and the reanalysis datasets, the variability and the circulation features influencing precipitation in the Tibetan Plateau (TP) are investigated. Taking into account the effects of topography, surface winds are deconstructed into flow-around and flow-over components relative to the TP. Climatologically, the flow-around component mainly represents cyclonic circulation in the TP during the summer. The transition zone of total precipitation in the summer parallels the convergence belt between the southerlies and the northerlies of the flow-over component. The leading mode of rainfall anomalies in the TP has a meridional dipole structure, and the first principal component (PC1) mainly depicts the variation of rainfall in the southern TP. The wet southern TP experiences strengthened flow-over, which in turn mechanistically favors intensified ascent forced by the flow-over component. In addition, variations in the Indian summer monsoon (ISM) have an important role in influencing the flow over the southern TP, and the ISM ultimately impacts the precipitation over southern TP.展开更多
文摘流量分类是优化网络服务质量的基础与关键.机器学习算法利用数据流统计特征分类流量,对于识别加密私有协议流量具有重要意义.然而,特征偏置和类别不平衡是基于机器学习的流量分类研究所面临的两大挑战.特征偏置是指一些数据流统计特征在提高部分应用识别准确率的同时也降低了另外一部分应用识别的准确率.类别不平衡是指机器学习流量分类器对样本数较少的应用识别的准确率较低.为解决上述问题,提出了基于集成聚类的流量分类架构(traffic classification framework based on ensemble clustering,简称TCFEC).TCFEC由多个基于不同特征子空间聚类的基分类器和一个最优决策部件构成,能够提高流量分类的准确率.具体而言,与传统的机器学习流量分类器相比,TCFEC的平均流准确率最高提升5%,字节准确率最高提升6%.
文摘In this paper we summarize the characteristics of the dishpan experiment, the principle of substance revolving, and the scientific basis of the “retrograde wave in only one direction” with respect to weather data and S. C. OuYang＇s articles in which the fundamental questions in the meteorological theory were pointed out. Furthermore, we discuss the systematic changes involving the concept, theory, and method that substance evolves.
基金supported by the National Basic Research Program of China (973 program, Grant No.2010CB950400)the National Natural Science Foundation of China (Grant No. 41030961)
文摘Based on the observation data and the reanalysis datasets, the variability and the circulation features influencing precipitation in the Tibetan Plateau (TP) are investigated. Taking into account the effects of topography, surface winds are deconstructed into flow-around and flow-over components relative to the TP. Climatologically, the flow-around component mainly represents cyclonic circulation in the TP during the summer. The transition zone of total precipitation in the summer parallels the convergence belt between the southerlies and the northerlies of the flow-over component. The leading mode of rainfall anomalies in the TP has a meridional dipole structure, and the first principal component (PC1) mainly depicts the variation of rainfall in the southern TP. The wet southern TP experiences strengthened flow-over, which in turn mechanistically favors intensified ascent forced by the flow-over component. In addition, variations in the Indian summer monsoon (ISM) have an important role in influencing the flow over the southern TP, and the ISM ultimately impacts the precipitation over southern TP.
