The program that infected Dolphin Stadium Web site visitors was seeking only logon credentials to World of Warcraft, rather than a myriad of other possible targets. The compromise of the Web site was not a blaring defacement, but aimed at infecting victims without attracting undue attention and timed to maximize its effectiveness. The presence of a keylogger on the infected computer was not easily noticeable, in contrast to attacks that infected machines with adware and until recently dominated the threat landscape.
Drive-by infections like the Dolphin Stadium incident that affect Web site visitors and install keyloggers are all too common. Although logon credentials to banking sites are still a popular target of keyloggers, usernames and passwords for other types of Web sites, such as gaming, social networking and job posting sites, are also targeted. While the immediate purpose of such campaigns is sometimes unclear, an interest in credentials for diverse categories of Web sites indicates a long-term effort by well-funded and dedicated attackers to assemble data warehouses.
Stealthy and Self-Preserving
The increasing profitability of targeted attacks has fueled investment in the development of malicious software that helps make them possible. Such efforts produce malware with capabilities that stretch our defensive abilities.
- Using protected tunnels and peer-to-peer protocols for malicious traffic is becoming increasingly common. There are bot specimens that employ SSL to encrypt their command and control (C&C) channels. Another approach has been employed by some Phatbot and SDBot variants, whose use of peer-to-peer protocols makes it particularly challenging to disrupt their C&C communications. A keylogger recently demonstrated another technique, in which it obfuscated its messages to the attacker and embedded them in Internet Control Message Protocol packets. Malware may also use the ubiquitous HTTP protocol when calling home, which helps it pass through firewalls and travel unnoticed in other Web traffic.
- Rootkits are getting better at concealing the presence of malware on the infected system and occur more frequently. Recently released Rustock and Unreal rootkits are highly effective at shielding themselves from common rootkit scanners that look for discrepancies in the infected system's configuration. Proof-of-concept rootkits such as SubVirt, Blue Pill and Vitriol are able to treat the infected system as a virtualized one, making their detection very challenging. Another emerging category of rootkits focuses concealment within applications. For instance, the Argeniss rootkit for Oracle can hide the attacker's database of choice--anything from collections of pornography to stolen credit card data--after the intrusion.
- The complexity of anti-analysis techniques employed by malware is continuing to evolve. Protecting malware from reverse engineering conceals the authors' plans and protects the C&C channel. Protective measures often involve detecting the presence of virtualization and debugging software commonly used by analysts. Rather than including such functionality directly in the malicious code, malware authors often rely on packers--programs that can add anti-analysis mechanisms to almost any executable. For example, Themida is a commercial packer highly effective at complicating malware analysis.
The Collective Power of BOTS
Stealth and self-preservation are characteristic of many types of malicious software; however, no malware is as prominent on the current threat landscape as bots.
A massive DDoS attack on the CastleCops Web site earlier this year demonstrated their power. A community of antimalware and antispam activists, CastleCops reported that the traffic directed at its Web site by the attacker's bots peaked at almost 1 Gbps. The flood inundated CastleCops' Internet pipe, making the Web site largely inaccessible for several days. The situation was reminiscent of DDoS attacks on spam-fighting sites Spamhaus, Spamnation and Blue Security; Blue Security ultimately could not withstand the sustained attack and closed down for good.
Such assaults demonstrate that attackers are becoming more aggressive at defending their spam- and malware-driven business models by punishing organizations and individuals they consider threats. Bots are a powerful weapon attackers can use for this purpose, providing attackers with the ability to command thousands of infected computers with a few keystrokes.
Being able to launch DDoS attacks allows the owner of the bot network (botnet) to discipline adversaries, extort money under the threat of such an attack, and offer attack services to others. One case of "DDoS for hire," documented by the FBI, involved the arrest of an owner of a sports apparel company on charges of hiring an attacker to disable the competitors' Web sites.
Sending spam messages is another common reason for employing a botnet. The spammer uses infected computers as spam relays, which provides him or her with virtually unlimited bandwidth and makes it very difficult for the defenders to block or trace the source of the unwanted messages.
Joe Stewart, a senior security researcher with managed security provider Secure-Works, in January analyzed a spam-related botnet powered by the Rustock Trojan that appeared to manipulate the price of a penny stock. The so-called pump-and-dump technique involved the attacker purchasing stock of a little-known company, hyping it up via spam messages and selling it as soon as it slightly increased in price. The technique can be surprisingly effective, earning a spammer as much as $20,000 over a weekend. Stewart's Web site was subsequently targeted by a DDoS attack the day after a newspaper described his analysis of the campaign.
Bot herders have been targeting desktop PCs of broadband users, using the infected computers as unwitting participants in a botnet, but this trend may be changing. Researchers at security assessment specialist Beyond Security--Gadi Evron, Kfir Damari and Noam Rathaus--this year observed an increase in the use of Web servers to construct botnets.
Server-focused bot herders exploit vulnerabilities in Web applications that are built using languages such as ASP, Perl and PHP to invoke their own scripts. Tynan Wilke, a SecureWorks researcher, documented one such campaign. In this attack, the bot herder used Google to locate servers with a vulnerable open-source Horde Webmail application and took advantage of the vulnerability to install a malicious Perl script. The bot allowed the attacker to launch DoS attacks, query Google for further propagation, and execute commands on the compromised Web server, according to Wilke.
Browser Malware
While botnets have become potent weapons for cyberattackers, online thieves are also turning their focus to the Web browser. The browser is becoming the primary application used to access data at home and at work, making it an attractive target. After all, why bother compromising the underlying operating system if the most sensitive transactions--from online banking to corporate sales management--occur in the browser? The browser includes powerful functionality to support the advanced requirements of modern Web-based applications; these features create an ecosystem for malicious code to survive without directly interacting with the operating system.
October 2005 brought the first high-profile worm that was purely Web-based. The Sammy worm took advantage of a cross-site scripting (XSS) flaw in the MySpace Web site and employed a popular JavaScript construct used in many AJAX applications. Such worms embed their code in pages of the compromised Web site and typically spread when the site's users view the infected page. The payload of such worms varies, but could range from defacing the infected pages to executing financial transactions within the context of the victim's session.
The Sammy worm infected more than a million MySpace users. On its heels came other worms powered by XSS and AJAX techniques. The list includes MySpace worms that propagated with the help of Flash and QuickTime browser plug-ins, as well as worms that spread on Orkut, Gaia Online and Yahoo! Mail Web sites.
Another example of the power of browser-based malware is the proof-of-concept port-scanning tool written in JavaScript by security vendor SPI Dynamics to demonstrate some of the challenges of securing intranets. Running in the victim's browser, the scanner can perform reconnaissance against the victim's network even if it is behind a firewall. This tool illustrates the extent to which the malicious Web site can explore the internal network of the site's visitor, even when operating purely in the browser.
SPI Dynamics further demonstrated the capabilities of browser malware by exhibiting a JavaScript bot at the ShmooCon conference in March without making its code public. Called Jikto, the proof-of-concept bot can locate vulnerabilities in Web applications while running within a Web browser. An attacker could inject Jikto in the victim's browser by exploiting XSS and other Web site vulnerabilities. The attacker could control Jikto instances remotely, capturing the information they collect and instructing them to launch further Web-based attacks.
Clearly, today's threat environment is multifaceted and rapidly changing. From zero-day exploits, client-side attacks and botnets, organizations are facing a maturing marketplace that encourages attackers to invest in better organization and tools. In turn, defenders need to keep learning from each other, sharing threat information and discussing effective defense strategies. This is the only way to ensure they do not fall behind in the cyberspace arms race that is unlikely to end any time soon.