System predicts 85% of cyber-attacks using input from human experts

Is it me or is the press release really bad? This publication seems to talk about some fraud prevention system rather than cyber attacks. Maybe fraud with a computer is now a cyber attack? The paper[1] cites this as malicious behavior (8.2):

• Terms of service abuse: This category covers fraudulent violations of the terms of service agreement. Frauds of this sort have very distinct signatures. Two simple examples are the abusive use of promotional codes, or deleting the web browser’s cookies to participate more times than allowed in an online voting platform.

Incognito-Mode is the way to be a cyber-attacker! The paper is really interesting and great, but I guess it has nothing to do with this headline:

> MIT reveals AI platform which detects 85 percent of cyberattacks

It's more like online-vote manipulation with while (1); do curl -XPOST <url>; done won't work anymore.

1: https://people.csail.mit.edu/kalyan/AI2_Paper.pdf

Cyber attacks seem to be employing a bit of AI and data science these days.

Url changed from http://www.zdnet.com/article/mit-reveals-ai-platform-which-d..., which copies this without linking to it. I didn't realize that zdnet had descended quite to that.