Check out the new USENIX Web site.


Introduction

It should come as no surprise that our increasing reliance on the Internet for many facets of our daily lives (e.g., commerce, communication, entertainment, etc.) makes the Internet an attractive target for a host of illicit activities. Indeed, over the past several years, Internet services have witnessed major disruptions from attacks, and the network itself is continually plagued with malfeasance [14]. While the monetary gains from the myriad of illicit behaviors being perpetrated today (e.g., phishing, spam) is just barely being understood [11], it is clear that there is a general shift in tactics--wide-scale attacks aimed at overwhelming computing resources are becoming less prevalent, and instead, traditional scanning attacks are being replaced by other mechanisms. Chief among these is the exploitation of the web, and the services built upon it, to distribute malware.

This change in the playing field is particularly alarming, because unlike traditional scanning attacks that use push-based infection to increase their population, web-based malware infection follows a pull-based model. For the most part, the techniques in use today for delivering web-malware can be divided into two main categories. In the first case, attackers use various social engineering techniques to entice the visitors of a website to download and run malware. The second, more devious case, involves the underhanded tactic of targeting various browser vulnerabilities to automatically download and run--i.e., unknowingly to the visitor--the binary upon visiting a website. When popular websites are exploited, the potential victim base from these so-called drive-by downloads can be far greater than other forms of exploitation because traditional defenses (e.g., firewalls, dynamic addressing, proxies) pose no barrier to infection. While social engineering may, in general, be an important malware spreading vector, in this work we restrict our focus and analysis to malware delivered via drive-by downloads.

Recently, Provos et al. [20] provided insights on this new phenomenon, and presented a cursory overview of web-based malware. Specifically, they described a number of server- and client-side exploitation techniques that are used to spread malware, and elucidated the mechanisms by which a successful exploitation chain can start and continue to the automatic installation of malware. In this paper, we present a detailed analysis of the malware serving infrastructure on the web using a large corpus of malicious URLs collected over a period of ten months. Using this data, we estimate the global prevalence of drive-by downloads, and identify several trends for different aspects of the web malware problem. Our results reveal an alarming contribution of Chinese-based web sites to the web malware problem: overall, 67% of the malware distribution servers and 64% of the web sites that link to them are located in China. These results raise serious question about the security practices employed by web site administrators.

Additionally, we study several properties of the malware serving infrastructure, and show that (for the most part) the malware serving networks are composed of tree-like structures with strong fan-in edges leading to the main malware distribution sites. These distribution sites normally deliver the malware to the victim after a number of indirection steps traversing a path on the distribution network tree. More interestingly, we show that several malware distribution networks have linkages that can be attributed to various relationships.

In general, the edges of these malware distribution networks represent the hop-points used to lure users to the malware distribution site. By investigating these edges, we reveal a number of causal relationships that eventually lead to browser exploitation. More troubling, we show that drive-by downloads are being induced by mechanisms beyond the conventional techniques of controlling the content of compromised websites. In particular, our results reveal that Ad serving networks are increasingly being used as hops in the malware serving chain. We attribute this increase to syndication, a common practice which allows advertisers to rent out part of their advertising space to other parties. These findings are problematic as they show that even protected web-servers can be used as vehicles for transferring malware. Additionally, we also show that contrary to common wisdom, the practice of following ``safe browsing'' habits (i.e., avoiding gray content) by itself is not an effective safeguard against exploitation.

The remainder of this paper is organized as follows. In Section 2, we provide background information on how vulnerable computer systems can be compromised solely by visiting a malicious web page. Section 3 gives an overview of our data collection infrastructure and in Section 4 we discuss the prevalence of malicious web sites on the Internet. In Section 5, we explore the mechanisms used to inject malicious content into web pages. We analyze several aspects of the web malware distribution networks in Section 6. In Section 7 we provide an overview of the impact of the installed malware on the infected system. Section 8 discusses implications of our results and Section 9 presents related work. Finally, we conclude in Section 10.

Niels Provos 2008-05-13