End-to-end Security for Web Applications : A Language-based Approach

133 views · Published 6 September 2016 · 1:08:08 · Indexed 10 November 2026

Channel: Microsoft Research · 2016 · Science & Technology

Watch on YouTube

Most large organizations must maintain a substantial information  presence on the world wide web in order to share information with  their partners and customers. For instance, the United States military  has begun using Intellipedia, a wiki-based online document management  system, in order to promote information sharing between the sixteen  agencies that comprise the U.S. intelligence community. Among other  security requirements, such an application should only allow  authorized users to access sensitive portions of a document, it should  track the provenance of data in each document and it should ensure  that information releases follow a specific downgrading  protocol. Web-specific threats, like script injection attacks, must  also be thwarted if critical data like authentication tokens are to be  protected. A framework that ensures that such a wide range of security  concerns is correctly addressed is highly desirable but, to date, no  such framework exists.    In this talk, I present SELinks, an extension of the Links programming  language in which web applications can be shown to correctly enforce a  wide variety of security policies end to end. In SELinks, a programmer  specifies a custom security policy by associating security labels with  sensitive operations and data. SELinks prevents a policy from being  circumvented by allowing labeled terms to be manipulated only within a  separate part of the program called the enforcement policy;  application code must treat labeled values abstractly. SELinks is also  equipped with support for policies that protect users running a  specially modified browser from script injection attacks. We have used  SELinks to build two substantial applications, including a secure  online document management system. Our initial experience indicates  that it is relatively easy to correctly enforce many common policies  in SELinks and, using a formal model, to prove that correct  enforcement entails the fulfillment of high-level security objectives.

More from this channel