社会化信息网络的内容分发需求感知与整形

The Internet, originally a technical network, has evolved into a platform of social activity. Under this socialization trend of information network, an acute prevalence of the distribution of large-volume content, such as high-definition video, has drown the network under tremendous pressure of bandwidth and storage. The users’ different content interests and the growth of the number of content resources lead to a long-tail distribution of content demands. The current network optimizing mechanism, in which the network is scheduled to adapt to the user demands, is facing great challenges. Moreover, there exists explicit and implicit social relationship between the Internet users who access to different contents, hence the user behaviors and content demands on the Internet becoming social significantly. Therefore, the traditional assumption of independent user behavior and independent resources is no more realistic. .To face the challenges mentioned above, from a social-network perspective, we propose the research on optimizing content delivery via shaping users content demands to fit for the network infrastructure without sacrifice the QoS guarantee. We aim at establish a novel network optimization framework and develop the network theory to adapt to the socialization trend in the information network. .Based on a large-scale commercial content distribution network，we intend to conduct this study in the following aspects with large-scale network measurement and data mining techniques: 1) Characterize the structures of the user social network and resource network implicit in the content accesses, examine the relationship between these structural properties and the scale of distinct content resource accesses, and then forecast the scales of content demands; 2)Based on the above studies, analyze and model the performance of content distribution network considering the correlation among users behavior and that among contents, and design practical network optimization algorithms thereafter; 3)Based on the demand prediction, explore the group recommendation mechanism over inhibit social network and the information cascades in explicit social networks, guiding users access behavior to shape the demands, i.e. to adjust the spatial and temporal distribution of users’ accesses to different contents, to follow the network capacity constraints on bandwidth and storage, and optimize the network resource utilization and the quality of service as well; and 4) Based on the demand shaping, design optimization mechanisms for network topology establishing and resources deployment.

信息网络的社会化演进，加速内容需求海量增长，导致现有网络优化框架难以在有限网络资源与不断增长内容需求的矛盾下提供令用户满意的服务。信息网络中用户之间的显式及隐式社会关系，呈现为用户兴趣及内容需求的显著相关性，动摇了经典网络服务质量理论中的用户行为独立性假设，但也提供了新的优化空间。本课题提出研究“社会化信息网络的内容需求感知与整形”，旨在建立一种基于需求感知与需求整形的网络优化新框架，探索网络理论在信息网络社会化演进中的新发展。拟基于自有平台及实际运营系统进行大规模网络测量与数据分析，部署优化机制并开展网络实验，研究基于显/隐式社会网络的信息网络内容需求感知，建立社会性内容需求与网络资源、服务质量的关系模型，提出基于隐式和显式社会网络的两类需求整形机制，在时间、空间的多方向、多位置上，引导用户需求适应网络能力，满足用户兴趣与服务质量要求，实现基于需求感知与整形的网络结构及资源部署的优化。

信息网络的社会化演进，加速内容需求海量增长，导致现有网络优化框架难以在有限网络资源与不断增长内容需求的矛盾下提供令用户满意的服务。信息网络中用户之间的显式及隐式社会关系，呈现为用户兴趣及内容需求的显著相关性，动摇了经典网络服务质量理论中的用户行为独立性假设，也提供了新的优化空间。为此本课题开展了“社会化信息网络的内容需求感知与整形”研究，旨在建立一种基于需求感知与需求整形的网络优化新框架，探索网络理论在信息网络社会化演进中的新发展。研究过程中课题组注意到，社会化信息网络服务中隐私风险与隐私保护成为用户关切，及其是否愿意选择网络服务的重要指标，也与本课题研究目标密切相关。因此课题组实际了开展以下三方面研究。1)需求感知与质量建模方面：完成了个体内容兴趣偏好需求和用户生命周期行为的分析模型与预测，对LTE环境点播视频服务、WiFi环境下实时视频服务和基于WiFi-Direct的LTE合作视频流媒体系统进行了QoE测量建模和预测，还进行了基于移动轨迹数据的需求空间分布感知，进一步提出了一种社会性内容需求与网络资源、服务质量的关系模型QI-E，为需求整形和服务优化提供理论支持。2)需求整形与服务优化方面：研究了网络开关流量需求整形、基于需求整形的视频数据传输机制优化、内容感知的QoE最优化网络资源分配、用户视角感知的沉浸式视频系统缓存与分发优化、基于需求整形的云下载服务动态资源分配、引入边缘计算能力的在线视频缓存，以及在线学习网络中用户状态和需求感知的习题推荐。3)隐私风险感知与保护方面：完成了视频观看行为隐私风险感知、个性化服务中跨系统隐私风险感知、个性化服务用户隐私保护算法、空间流数据发布中的轨迹隐私保护算法、基于深度学习和增强学习的智能认证方法。基于上述研究，我们探索了社会化信息网络中需求、体验与资源三要素关系的理论模型，形成了一系列需求感知与需求整形理论模型与优化方法，为社会化信息网络的优化提供了理论支持，具有广泛应用潜力。