The social networks have been constantly targeted by attackers for stealing users’ accounts, credit card info, sending spam, etc. The attackers have all kinds of resources they need in order to launch serious attacks – sufficient funds, cheap labor, infected bot machines and compromised accounts. Protecting users in social networks is a very challenging adversarial learning problem. One must consider both the product needs and the scale of the users in social networks. In the past several years, Twitter has set up a universal and scalable real-time safety system to protect its users and their social graph.
Twitter’s detection and defense system will check every read and write actions of every user in real time, in order to classify and examine its validity. Nowadays there are about 8,000 tweets every second – around 700 million tweets per day, which leads to more than 100 billion checks per day by this safety system. This system is the foundation of Twitter’s security architecture. Based on this system, we are able to implement many applications such as fake accounts detection and automation detection or even block iOS and Android app bugs. Many signals and features produced by this system can also be further utilized by machine learning and other advanced analysis systems. This system was originally designed by Facebook’s safety team, and then widely adopted by other Silicon Valley companies such as Snapchat, Airbnb, Lyft, LinkedIn and Tencent’s Wechat.
Dr. Jin Han is the co-founder and CTO of a Silicon Valley based blockchain startup. As the former engineering manager and full stack engineer of Twitter’s security team, Jin has more than a decade of experience in conducting security research and providing industrial security solutions. Jin also serves as the security advisor of different blockchain companies including exchanges, wallets and mining companies. Jin has bachelor’s degrees from both Fudan University and University College Dublin, Ireland. He obtained his master’s degree from Fudan university and his PhD from Singapore Management University.
With the efforts of Jin’s security team, Twitter’s real-time detection and defense system has been a huge success, which is able to reduce more than 95% of Twitter’s spam and abuse user reports and turned around Twitter’s public image on safety and abuse problems.