Using Dynamic Analysis and Template Matching for Phishing Detection (waset.org)

I was just thinking about whether an approach like this could work, so I did some searching and found some research. There's also [Detection of Phishing Webpages based on Visual Similarity](http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.59....).

Does anyone know if such a system is getting worked on, say, at Google?