AI

ZARATHUSTRA: Extracting WebInject Signatures from Banking Trojans

Abstract

Modern trojans are equipped with a functionality, called WebInject, that can be used to silently modify a web page on the infected end host. Given its flexibility, WebInject-based malware is becoming a popular information-stealing mechanism.

In addition, the structured and well-organized malware-as-a-service model makes revenue out of customization kits, which in turns leads to high volumes of binary variants. Analysis approaches based on memory carving to extract the decrypted webinject.txt and config.bin files at runtime make the strong assumption that the malware will never change the way such files are handled internally, and therefore are not future proof by design. In addition, developers of sensitive web applications (e.g., online banking) have no tools that they can possibly use to even mitigate the effect of WebInjects.

WebInject-based trojans insert client-side code (e.g., HTML, JavaScript) while the targeted web pages (e.g., online banking website, search engine) are rendered on the browser. This additional code will capture sensitive information entered by the victim (e.g., one-time passwords) or perform other nefarious actions (e.g., click fraud or search engine result poisoning). The visible effect of a WebInject is that a web page rendered on infected clients differs from the very same page rendered on clean machines. We leverage this key observation and propose an approach to automatically characterize the WebInject behavior. Ultimately, our system can be applied to analyze a sample automatically against a set of target websites, without requiring any manual action, or to generate fingerprints that are useful to determine whether a client is infected. Differently from the state of the art, our method works regardless of how the WebInject module is implemented and requires no reverse engineering.

We implemented and evaluated our approach against live online websites and a dataset of distinct variants of WebInject-based financial trojans. The results show that our approach correctly recognize known variants of WebInject-based malware with negligible false positives. Throughout the paper, we describe some use cases that describe how our method can be applied in practice