Opera Backstage: Opera 9.1 will include Fraud Protection
By Johan Borgborg. Tuesday, October 17, 2006 1:11:05 PM
When you go to a new site for the first time, Opera will check against a database if the site is trusted or if it is a known fraud site. If we know the site, there will be a small information "i" in the right end of the address field. If it's unknown/not verified there will be a "?" and if it's known as a fraudulent site we will display a warning and block the user from accessing the site.
The browser sends only the minimum information the database needs to identify a fraud site. When a result is received by the browser, it will be cached there for some time, so it doesn't have to check again if you go to the same site often.
- Why don't we use a downloaded blacklist like Firefox 2?
Firefox 2 only checks against a blacklist unless you turn on real-time protection from Google or other providers. We feel that only real-time protection is real protection, since phishing attacks tend to be more and more like virus attacks, most of their damage is done in a very short time.
- Why don't we use a downloaded whitelist like IE 7?
This makes some sense, especially to save bandwidth for our servers. But for the privacy-concerned user, we don't think it changes anything, since it's typically the more obscure sites that you really want to keep to yourself. We've made it easy to turn on and off the fraud protection from the information dialog you get when clicking the icon.
More technical details:
When you browse to a site you have not visited before, the browser sends a request for site information to our server. The requests contains the domain name of the site and a hash value of the URL. We don't send the full URL, but we need a fingerprint of the full URL in case you visit a dangerous page on a site that is otherwise harmless.
The reply from the server is an XML document containing the trust level of the domain. This reply will be cached by Opera for a time indicated by our server. This means that information about well-trusted sites can be cached for a longer period than for unknown sites.
The requests go over HTTP, but the replies will be signed by the server to make sure they are genuine. We prefer to send information between the browser and ourselves in plain text, so our users can inspect the data we send "home".