What Is Phishing (Or, How to Fight Phishing at the User-Interface Level)
Pages: 1, 2, 3, 4, 5, 6

Phishing as a Semantic Attack

Bruce Schneier has observed that methods for attacking computer networks can be categorized in waves of increasing sophistication and abstraction. According to Schneier, the first wave of attacks was physical in nature, targeting the computers, the network devices, and the wires between them, in order to disrupt the flow of information. The second wave consisted of syntactic attacks, which target vulnerabilities in network protocols, encryption algorithms, or software implementations. Syntactic attacks have been a primary concern

figure 2
Figure 2. Screenshot of a phishing web page pointed to by the phishing email (source: Anti-Phishing Working Group)

figure 3a
figure 3b
Figure 3. Anatomy of a phishing attack

of security research for the last decade. The third wave is semantic: "attacks that target the way we, as humans, assign meaning to content." [5]

Phishing is a semantic attack. Successful phishing depends on a discrepancy between the way a user perceives a communication, like an email message or a web page, and the actual effect of the communication. Figure 4 shows the structure of a typical Internet communication, dividing it into two parts. The system model is concerned with how computers exchange bits—protocols, representations, and software. When human users play a role in the communication, however, understanding and protecting the system model is not enough, because the real message communicated depends not on the bits exchanged but on the semantic meanings that are derived from the bits. This semantic layer is the user's mental model. The effectiveness of phishing indicates that human users do not always assign the proper semantic meaning to their online interactions.

figure 4
Figure 4. Human-Internet communication

When a user faces a phishing attack, the user's mental model about the interaction disagrees with the system model. For example, the user's intention may be "go to eBay," but the actual implementation of the hyperlink may be "go to a server in South Korea." It is this discrepancy that enables the attack, and it is this discrepancy that makes phishing attacks very hard to defend against. Users derive their mental models of the interaction from the presentation of the interaction—the way it appears on the screen. The implementation details of web pages and email messages are hidden, and are generally inaccessible to most users. Thus, the user is in no position to compare his mental model with the system model, and it would take extra effort to do so. On the other hand, email clients and web browsers follow the coded instructions provided to them in the message, but are unable to check the user's intentions. Without awareness of both models, neither the user nor the computer is able to detect the discrepancy introduced by phishing.

One extreme solution to the phishing problem would simply discard the presentation part of an Internet communication—the part that produces the user's mental model—because it can't be trusted. Instead, a new presentation would be generated directly from the implementation. If the user's computer is trustworthy, then, the presentation seen by the user would be guaranteed to be related to the actual implementation. Unfortunately, the cost of this idea in both usability and functionality would be enormous. Most online messages are legitimate, after all, with the presentation correctly reflecting the implementation. Phishing messages are rare (but pernicious) exceptions. So this solution would improperly sacrifice the freedom of legitimate senders to present and brand themselves in order to block a small number of wrongdoers.

So we must accept the fact that users will see messages with mismatched presentation and implementation. Attempts to fight phishing computationally, which are discussed in this chapter, try to enable the computer to bridge the gap between the user's mental model and the true system model. But the human user must be the final decision-maker about whether a message is phishing. The reason is that phishing targets how users assign semantic meaning to their online interactions, and this assignment process is outside the system's control.

Pages: 1, 2, 3, 4, 5, 6

Next Pagearrow