网络流量异常检测中的维数约简研究

对包含大流量数据的高维度网络进行异常检测,必须加入维数约简处理以减轻系统在传输和存储方面的压力。介绍高速网络环境下网络流量异常检测过程以及维数约简方式,阐述流量数据常用特征和维数约简技术研究的最新进展。针对网络流量特征选择和流量特征提取2种特征降维方式,对现有算法进行归纳分类,分别描述算法原理及优缺点。此外,给出维数约简常用的数据集和评价指标,分析网络流量异常检测中维数约简技术研究面临的挑战,并对未来发展方向进行展望。

网络加密流量识别研究进展及发展趋势

网络加密流量的快速增长正在改变威胁形势。如何实现对网络加密流量的实时准确识别,是我国网络空间安全领域的重要问题,也是目前网络行为分析、网络规划建设、网络异常检测和网络流量模型研究的重点。文章对网络加密流量识别的基本概念、研究进展、评价指标和存在的问题进行论述,并对网络加密流量识别的发展趋势和面临的挑战进行总结与展望。文章可为进一步探索网络空间安全领域的新方法与新技术提供借鉴与参考。

网络攻击检测中流量数据抽样技术研究

网络攻击检测的关键是通过对网络流量的分析快速识别网络中的未知攻击行为。大数据环境下,如何在不影响异常检测效果的前提下通过数据抽样技术粗粒度减少需要处理的网络流量数据,筛选出需要进行细粒度异常检测的子集,为网络攻击发现提供可靠数据支撑,是网络入侵检测系统研究的重要问题,也是目前网络行为分析、网络测量分析、网络异常检测、网络流量模型研究的重点。文章对网络攻击检测中流量数据抽样技术的基本概念、研究进展和存在问题进行阐述,对网络流量数据抽样技术面临的挑战和发展趋势进行总结和展望。文章可为进一步探索网络攻击检测领域的新方法和新技术提供借鉴和参考。

面向小样本数据的机器学习方法研究综述

小样本学习是面向小样本数据的机器学习,旨在利用较少的有监督样本数据去构建能够解决实际问题的机器学习模型。小样本学习能够解决传统机器学习方法在样本数据不充分时性能严重下降的问题,可以为新型小样本任务实现低成本和快速的模型部署,缩小人类智能与人工智能之间的距离,对推动发展通用型人工智能具有重要意义。从小样本学习的概念、基础模型和实际应用入手,系统梳理当前小样本学习的相关工作,将小样本学习方法分类为基于模型微调、基于数据增强、基于度量学习和基于元学习,并具体阐述这4大类方法的核心思想、基本模型、细分领域和最新研究进展,以及每一类方法在科学研究或实际应用中存在的问题,总结目前小样本学习研究的常用数据集和评价指标,整理基于部分典型小样本学习方法在Omniglot和Mini-ImageNet数据集上的实验结果。最后对各种小样本学习方法及其优缺点进行总结,分别从数据层面、理论研究和应用研究3个方面对小样本学习的未来研究方向进行展望。

Structural Phase Transitions of Iodine at High Pressure

The compression and polymorphic phase transitions of iodine have been studied by using x-ray diffraction with a diamond anvil cell at high pressure up to 77GPa.Considering that iodine is reactive with iron in air,a corrosion-resisting and high-strength gasket was put between the diamonds,thus the sample can be mounted at atmospheric pressure and room temperature.Pressure was measured in terms of the ruby fluorescence and the molybdenum diffraction as an internal pressure standard.According to the x-ray diffraction results,the iodine undergoes the following three structural phase transitions with pressure increasing:(1)At 21 GPa,from base centred orthorhombic phase(phase Ⅰ)to body centred orthorhombic phase(phase Ⅱ).A 2%change in volume occurs.(2)At 44 GPa,from phase II to body centred tetragonal phase(phase Ⅲ).This is a second order phase transition.(3)At 71 GPa,from phase III to fcc phase(phase Ⅳ).The volume decreases by 3%.At this time,the volume is 45%of the original one at atmospheric pressure.

Solid C_(60):Pressure Induced Phase Transitions

Based on in situ high pressure x-ray diffraction with diamond anvil cell by using synchrotron radiation and pressure-quench x-ray measurement,phase transition of C_(60)under high pressured is studied.The maximum pressure used is 30GPa in quasi-hydrostatic case but 44GPa in nonhydrostatic case.At ambient pressure,C_(60)crystal is a fcc phase with lattice parameter a=14.2086Å.It transforms from fcc to a lower symmetry phase at pressure of 13.7GPa,and then transforms to an amorphous state under pressure up to 25-30 GPa.This amorphous state is metastable at normal condition.Finally,the experimental results are discussed.

High pressure effects on the crystal structure andelectric conductivity of perovskite like (Ca/Sr)_2CuO_2Cl_2 compounds

The crystal structure stability as well as electric conductance of (Sr/Ca)_2CuO_2Cl_2 compound with K_2NiF_4 structure was investigated up to 31 GPa using diamond anvil technique. It seems that (Sr/Ca)_2CuO_2Cl_2 is quite stable under pressure but with obvious anisotropic compressibility. The equation of state (EOS) obtained shows relative large bulk modulus.

Synthesis of boron-doped diamond with laser heated diamond anvil cell

Boron-doped diamond has been synthesized from graphite mixed with different ratio of B 4 C at high pressure high temperature (HPHT) using laser heated diamond anvil cell.The starting composition was transformed to diamond compound at pressure ～9 GPa,2300-2400 K as indicated by the in-situ X-ray diffraction pattern with synchrotron radiation source.Raman spectrum of the recovered specimen from HPHT state confirmed that boron has been doped into diamond lattice.