Abstract. Security auditing of industry-scale software systems mandates automation. Static taint analysis enables deep and exhaustive tracking of suspicious data flows for detection of potential leakage and integrity violations, such as cross-site scripting (XSS), SQL injection (SQLi) and log forging. Research in this area has taken two directions: program slicing and type systems. Both of these approaches suffer from a high rate of false findings, which limits the usability of analysis tools based on these techniques. Attempts to reduce the number of false findings have resulted in analyses that are either (i) unsound, suffering from the dual problem of false negatives, or (ii) too expensive due to their high precision, thereby failing to scale to real-world applications.In this paper, we investigate a novel approach for enabling precise yet scalable static taint analysis. The key observation informing our approach is that taint analysis is a demand-driven problem, which enables lazy computation of vulnerable information flows, instead of eagerly computing a complete data-flow solution, which is the reason for the traditional dichotomy between scalability and precision. We have implemented our approach in ANDROMEDA, an analysis tool that computes data-flow propagations on demand, in an efficient and accurate manner, and additionally features incremental analysis capabilities. ANDROMEDA is currently in use in a commercial product. It supports applications written in Java, .NET and JavaScript. Our extensive evaluation of ANDROMEDA on a suite of 16 production-level benchmarks shows ANDROMEDA to achieve high accuracy and compare favorably to a state-of-the-art tool that trades soundness for precision.

JavaScript is the most popular client-side scripting language for Web applications. Exploitable JavaScript code exposes end users to integrity and confidentiality violations. Client-side vulnerabilities can cost an enterprise money and reputation, and cause serious damage to innocent users of the Web application. In spite of all this, recent research in the area of information-flow security has focused more on other languages that are more suitable for server-side programming, such as Java.Static analysis of JavaScript code is very challenging due to the dynamic nature of the language. This paper presents ACTARUS, a novel, product-quality static taint analysis for JavaScript that scales to large programs and soundly models all the JavaScript constructs with the exception of reflective calls. This paper discusses the experimental results obtained by running ACTARUS on a collection of 9,726 Web pages obtained by crawling the 50 most visited Web sites worldwide as well as 19 other popular Web sites. The results expose 526 vulnerabilities in 11 sites. Those vulnerabilities, if exploited, can allow malicious JavaScript code execution.