The technique of phase-space analysis is useful in fields that go far beyond sequence-number generation. Other parameters that are chosen pseudo-randomly or according to some internal scheme—such as IP packet ID fields, DNS request identifiers (as shown in Figure 10-16), application-generated “secret” cookies that identify user sessions, and so on—can be analyzed successfully, either to find flaws in a design or to identify an implementation and simplify further analysis or facilitate an attack.
Some work in this direction had been done or is under way; in a paper partly related to my original research, Joe Steward provides insight into some of the DNS system problems[88] that arise with the advancement of sequence number-prediction mechanisms. He notes that not only a UDP-based DNS protocol offers request verification methods that are simply not enough to withstand even “low-budget” spoofing attacks, but also the low quality of unique request identifiers generated by various implementations further weaken the scheme to make it trivially vulnerable to malicious data injection. Given that DNS is one of core services of the Internet, and that the perspective of spoofing a DNS response for a popular site to redirect all users of a specific network to a different web page is not exactly not tempting, DNS poisoning tops my list of downplayed threats on the Internet.
Dan Kaminsky provides some interesting, more advanced visualizations of supposedly random data at http://www.doxpara.com/pics/index.php?album=phentropy (Figure 10-17), definitely a worthy read.
