We are at the beginning of a fundamental shift in how content is created and exchanged over the Internet.  While content was previously created primarily by a small minority of organizations, now, individual users--empowered by the popularity of digital devices and social networks, as well as the ubiquity of Internet access--are creating content that represents a significant fraction of Internet traffic. Unfortunately, existing techniques and infrastructure are ill-suited for the new patterns of content creation and exchange, resulting in a mismatch of infrastructure and workload that is evident in places ranging from the ways in which content is distributed to the ways in users are able to express access control.  To make matters worse, existing providers have been slow to develop new techniques, as their current business models are often heavily reliant on existing approaches. 

Motivated by these trends, this project conducted measurement studies and developed systems, networks, and distribution architectures motivated by the changing patterns of content creation and exchange.  The goal of all of these systems is to enable users to freely exchange content and express meaningful privacy policies for end-user-generated content.

Intellectual Merit:

The first major component of this project was better understanding the evolution online social networking sites, as most prior studies have used data from only a small window of time.  Using a set of over 37 billion tweets spanning between 2006 and 2013, we demonstrated how the users, their behavior, and the site as a whole have evolved, quantifying a number of trends: the spread of Twitter across the globe, the percentage of tweets that are no longer available, the increasing fraction of Twitter user accounts that are inactive, and the increase of malicious behavior on Twitter. 

The second major component focuses on the sensitive content that is often shared with just subsets of friends. Due to the privacy-sensitive nature of the content, one of the hardest parts of using today’s OSN privacy management tools is defining appropriate groups for different pieces of content.  Using a popular Facebook application installed by over 1,000 users, we collected a total of 7,602 unique groups manually specified by users. We found that over 67% of users are sharing at least some of their uploaded content using user-defined subsets, and that 17.6% of all content is shared with such a subset of friends; these observations underscore the important and unstudied role that SACLs play in users’ privacy management.

The final major component of this project involved building Maygh, a system that allows direct exchange of static content between web browsers without the need for additional plug-ins or client-site software. Clients visiting the operator's web site automatically assist in serving static content to others for the duration of their visit; once they leave the site, this assistance ceases. As a result, Maygh allows operators who wish to distribute significant amounts of static content to spread the costs of distribution across their users.

Broader Impact:

The first two components offer a unique look into the large-scale evolution of an online social networking site and the fine-grained privacy settings that users choose when they share content.  Each of these has the potential to lead to a better understanding on how online social networks are causing human behavior and society as a whole to evolve. 

The Maygh project is the first system that can build a functional content distribution network from the visitors to a web site without requiring users to have installed a browser plug-in or other software. Thus, Maygh can directly be deployed to numerous sites today, and can lower the costs required to distribute static content.

Last Modified: 09/01/2016
Modified by: Alan Mislove