Identifying anomaly multimedia traffic in cyberspace is a big challenge in distributed service systems, multiple generation networks and future internet of everything. This letter explores meta-generalization for a multiparty privacy learning model in graynet to improve the performance of anomaly multimedia traffic identification. The multiparty privacy learning model in graynet is a globally shared model that is partitioned, distributed and trained by exchanging multiparty parameters updates with preserving private data. The meta-generalization refers to discovering the inherent attributes of a learning model to reduce its generalization error. In experiments, three meta-generalization principles are tested as follows. The generalization error of the multiparty privacy learning model in graynet is reduced by changing the dimension of byte-level imbedding. Following that, the error is reduced by adapting the depth for extracting packet-level features. Finally, the error is reduced by adjusting the size of support set for preprocessing traffic-level data. Experimental results demonstrate that the proposal outperforms the state-of-the-art learning models for identifying anomaly multimedia traffic.

翻译：在分布式服务系统、多代网络和未来的一切互联网中,发现网络空间的反常多媒体流量是一项巨大的挑战。本信探索了灰网多党隐私学习模式的超常规化,以改善反常多媒体流量识别的性能。灰网多党隐私学习模式是一个全球共享的模式,通过与保存的私人数据交换多党参数更新来分割、传播和培训。元化是指发现学习模式的固有属性,以减少其普遍性错误。在实验中,测试了三个元化原则如下。灰网多党隐私学习模式的普及性错误通过改变字面层嵌入的层面而减少。随后,错误通过调整提取包级特征的深度而减少。最后,错误通过调整预处理流量数据的支持规模而缩小。实验结果显示,该提案超越了用于识别异常多媒体流量的州级学习模式。