Machine Learning in Security and Integrity：With the emerging of e-market (e.g. Taobao, Amazon), cloud (e.g. Baidu Yun, AWS, Azure), and social network (e.g. WeChat, Facebook), the world becomes smaller and smaller with unbelievable easily running business, accessing large-scale resources and services, building community and sharing your moments cross continents within seconds. However, the concerns on security and integrity are arising too. Facing daily big data (petabyte or even more) with both imbalance and ambiguity, the real-time machine learning solution becomes challenging but demanding, which stimulates many breakthroughs.

This presentation starts with a quick overview of security problems, then it will discuss the machine learning platform, and focus on the topics related to the advanced technologies and solutions for the challenges facing in our time and the lessons from real applications. The presentation will end with the specific examples of recent machine learning applications in security.

演讲提纲

1. Overview of some security problems at the emerging e-market, cloud and social network 
   a. Transaction fraud and abuse 
   b. Cloud security 
   c. Integrity
2. Machine learning platform and structure 
   a. Offline build, online deployment, feedback loop 
   b. Data and feature engineering 
   c. Horizontal v.s. Vertical model structure  
   d. Traditional classification v.s. deep learning. 
3. Challenges and lessons learnt 
   a. Imbalanced data 
   b. Dealing with ambiguity: semi-label and no-label 
   c. Cold start problem
4. Application #1: Machine Learning in transaction risk management
5. Application #2: Machine Learning in cloud security

听众受益

With the talker’s numerous years of experience in the high-profile companies and dealing with the real-time solution for large world-wide data, the audience will learn from his experience (both success and lessons from failure) and perspective.

Audience will learn:

1. An overview of machine learning platform and structure, especially in security solution.
2. Through the two applications, audience will gain a valuable perspective from how advanced technology dealing with large scale data are used in the real-time system.

参考翻译：

随着电子商务如Taobao, Amazon的崛起，云计算如AliCloud, AWS, Azure的爆发，以及社交网络如WeChat, Facebook的大面积普及化，使得商业业务创新和开展越来越容易，轻而易举使用大规模资源和服务，全球化信息互动和共享已经可以同步，总的来说世界正在变得越来越小。但是，信息安全和商业诚信问题却不断增加，面对每天超过PB级的不平衡和模糊大数据，实时机器学习和解决方案面临着巨大的挑战。这样的要求在各个领域越来越多，促发了许多突破性的研究成果和应用。

演讲过程中会快速浏览一些安全问题，然后讲述一下机器学习平台，重点讲解能够处理当下棘手问题的相关先进技术和解决方案，以及一些真实的应用效果。最后会分享几个最近机器学习在安全应用上的特别案例。

演讲提纲

1. 总览几个电商、云计算和社交网络出现的安全问题 
   a. 交易欺诈和滥用 
   b. 云计算安全 
   c. 广告及其他商业诚信
2. 机器学习平台和架构 
   a. 离线构建，在线部署，反馈机制 
   b. 数据和特征工程 
   c. 平行 v.s 垂直模型结构 
   d. 传统分类 v.s 深度学习
3. 挑战和切身教训 
   a. 非平衡数据 
   b. 处理模糊性：半标签和无标签 
   c. 冷启动问题
4. 应用 1: 机器学习在交易风险管理方面的实践
5. 应用 2: 机器学习在云安全方面的实践

听众受益

讲师希望分享他十余年在IT界高端的几个公司里的宝贵经验。这些经验来自于对实时海量数据的机器学习及解决方案，成功和经验。

1. 整体了解机器学习平台和构架，尤其是在安全诚信领域中。
2. 通过分享的这两个应用，参会者能够进一步学习实时系统是如何使用先进的机器学习处理大规模数据。