Advertising poses more risks than those already discussed regarding AdWords, AdSense, and DoubleClick. Attackers can exploit advertising networks to compromise end-user machines, unethical interface techniques can trick users into disclosing sensitive information, and historically unbiased network providers can insert advertisements as the web pages make their way to the user’s browser.
Malicious Ad Serving
Advertising networks are more than just information-disclosure risks. They also serve as a malware attack vector. Advertising services pay web site owners for publishing advertisements on their web sites. A very common technique is the banner ad we’ve all seen at the top of web pages. Such ads usually take the form of animated GIF files, but they now include many image and video formats. Individuals and organizations that want to advertise using such a service create a media file and pay an advertiser a fee, and the advertiser serves the image to thousands of visitors of sites that belong to its advertising network. The risks here are twofold. Attackers have created misleading advertisements as a means to draw traffic to a malware serving or other malicious web site.47 The users’ trust of the advertisement company and the hosting web site increases their trust of the advertisements, leaving web surfers more vulnerable to such an attack. Virus writers have used the Google Adwords service to serve text ads that appeared to link to legitimate destination sites, but silently infected vulnerable web surfers by routing users through an intermediate, malicious site. Attackers also have used a vulnerability in Internet Explorer to compromise visitors as they passed through the intermediate site, before ultimately arriving at the legitimate site.48
The ads themselves have also been used to attack the web user directly. Malformed graphical images are one common technique. For example, a banner advertisement displayed on MySpace served spyware to more than one million visitors. In this case, attackers exploited a flaw in the way Windows processed Windows Meta File (WMF) images to install a Trojan horse.49 Because the attack occurs when the browser displays the image, the user needn’t click the advertisement to be infected. Another attack, served by DoubleClick, used rich media advertisements created in Adobe Flash to exploit a similar vulnerability, with the malicious advertisements appearing on extremely popular sites, including those of The Economist and Major League Baseball.50 Rich media advertisements are highly interactive and are becoming increasingly popular. Even the seemingly simple task of securing browsers against malicious images is proving difficult, and complex, rich media tools such as Flash are proving to be an even greater challenge.51 Other complex environments, such as ads embedded in Adobe PDF files, are now being explored as ways to reach potential customers, and I expect that similar issues will arise.52
Beyond malicious ad serving, advertisers employ another concerning strategy that I call malicious interfaces. In the idealistic world of interface design theory, interface designers always operate in the best interests of their users. Designers carefully study user tasks and painstakingly craft interfaces and applications to help users accomplish them. However, in the world of online advertising, the exact opposite is true. Designers frequently violate design best practices to coerce or mislead users into viewing advertising. Examples abound on the web: fake hyperlinks that pop up advertisements, giant advertisements that cover the text of articles, distracting advertising videos that begin playing the moment a page is viewed, banner ads with fake buttons that appear to be a part of the interface, advertisements embedded in video clips ... the list goes on. Malicious interface designers are creative; new “innovations” are coming out regularly. The only constraining factor appears to be the tolerance of the user. The invasiveness of advertisements is getting worse. You may have heard the term “banner blindness.” Users quickly learned that banner ads are of little value and ignore the advertisements, to the point that they barely perceive banner ads anymore. This defense mechanism has forced advertisers to become more aggressive in capturing user attention.54, 55 Although, I don’t claim that Google employs malicious interface design, the trend is concerning; it seems that malicious interface designers are carefully seeking the sweet spot between making advertising profit and annoying the user so much that they abandon a given site altogether.
As carriers of key components of network infrastructure, ISPs and web hosting services are flexing their muscle to place advertisements in front of users. For example, domain registrar and web hosting provider Network Solutions hijacked customers’ unused subdomains to resort to ad-laden “parking” pages.56 As another example, bloggers Lauren Weinstein and Sarah Lai Stirland reported Canadian ISP Rogers modification of web pages en route (see Figure 7-7).
Figure 7-7 Screenshot of an ISP altering a Google web page
Inserting advertisements into web pages as they transit an ISP is potentially very big business. ISPs already have access to a tremendous amount of personal information about users’ online activity. This data is a veritable gold mine if used to target online advertising. Some ISPs are reportedly selling significant amounts of user data to online marketers.57 In the United Kingdom, three major ISPs have announced plans to use user clickstream data to insert relevant advertisements as they surf, through a new startup called Phorm.58 ISP data contains some of the most sensitive information disclosures made by online users. If this advertising technique becomes widespread, virtually every web surfer’s activities will be passed on to advertisers in some form. Fighting the issue will be difficult, and users might find that they are faced with little alternative than to accept this new status quo.
Affiliate advertising isn’t just confined to Google via its AdSense and DoubleClick programs. It is a popular marketing practice in which, in its common form, web authors embed advertisements that contain unique tracking data to identify the correct affiliate to compensate. Such advertisements can be static—that is, the advertisement exists entirely on the server of the web author. In this case, the user must click the advertisement before the advertiser is aware of the user. However, many affiliate services provide dynamic content that is pulled directly from the online company without any action by the user, immediately linking the user to the visited web site (see Figure 7-8 for an example). In addition, the online company might tag the user with a cookie or retrieve an existing cookie, opening the possibility of identifying the user by name, billing address, and shipping address.
Figure 7-8 Example of Amazon affiliate network dynamic advertisements. Some affiliate networks, such as Amazon, encourage web authors to embed these advertisements into their web content, allowing the user to be logged, tagged with a cookie, and perhaps identified by name, merely by visiting the web page containing the advertisement.59
Facebook’s Beacon illustrates both the detailed insight that large online companies have into the activities of their user populations and the lengths some companies will go to increase profit. Facebook’s Beacon service allows Facebook users to share their purchases from affiliated online companies, such as books, movies, and gifts, with their Facebook friends.60 However, when the service was first offered, participation was “opt out.” The end result is that many users were infuriated and the civil action group MoveOn.org initiated a very effective online petition that rapidly gained more than 50,000 supporters.61 MoveOn even blamed Facebook for “ruining Christmas” because the Beacon advertising system allowed users to see holiday present purchases made by friends and family.62
Shortly thereafter, Facebook changed the service to “opt in” by requiring explicit user permission before publishing purchases to the user’s Facebook friends.