While there were many steps in the attacks, they began with bait messages, specifically because they are so efficient. By definition, if one puts bait before a sufficient number of people, someone will take it. The interesting thing is how small that number has to be.
The bad news is that only one click by one user may be sufficient to contaminate the entire enterprise. The good news is that all most all attacks against enterprises are starting in the same way.
The bait of choice no longer appeals to fear, greed, or lust. Rather it appeals to curiosity. Human beings are naturally curious; curiosity has survival value. Mass bait will be of the form "Look what Justin Bieber did." Alternately it may exploit the disaster news of the day. However, messages directed to the enterprise, while still appealing to curiosity, are much more artfully crafted. For example, the bait that compromised RSA was a pdf identified as "2011 Recruitment Plan." If this came to you from someone whose name you recognized, would you be suspicious? Would you resist it? Remember when we preferred PDFs to Word documents for safety?
The obvious defense against bait attacks is awareness training. However, as with campaigns like "Just Say No." there are fundamental limits to the effectiveness of such training. We are left with the fact that a successful attack only requires one temporary failure of our training.
I met Brian later and we agreed that we really need an effective and efficient artificial intelligence, AI, for identifying such messages. We both identify and reject one or two bait messages a day that get past our spam filters. If we can identify them, surely Google could.
However, I heard another presentation by Steve Ward, Vice President of Marketing for Invincea, speaking at Data Connectors at Bridgewater's at the end of Fulton Street. He talked about a product that took a different approach. It looked at the second step in the attack. It seems that one bites, i.e., "takes the bait," by clicking on a button. It turns out that almost all of the buttons are URLs. Steve says, even if I cannot stop everyone from biting, one might be able to cut lhe line just as they do. Only rare messages are bait but all bait messages are URLs.
The URLs link to an executable that corrupts the user's system. It effectively contaminates the network, all machines to which that machine is peer connected. In far too many enterprises, that is the entire enterprise network.
Note that contaminiation requires user privileges, perhaps ADMIN, at least the ability to create or modify an executable. Part of the problem is that users that do not require such privileges have them by default. On the other hand, we cannot limit all such privileges.
However, Steve Ward points out that controlling the process that parses the URL could prevent the contamination. His product takes an architectural approach, it installs as an application, becomes the parser for all URLs, and interprets them in a virtual machine so as to prevent contamination of the real machine. Even if a privileged user takes the bait, her machine will not be contaminated.
Efficient security relies upon layers and redundant measures. We must train users to recognize and resist bait. We must limit their privileges. We must configure their systems to resist contamination. We must layer and compartment the enterprise network to resist the spread of contamination. We must control access to sensitive data. We must monitor, detect and remediate. We must resist exfiltration of our data. Of course, it is because knowing and doing this is difficult that we are called professionals and are paid the big bucks.