http://www.phishtank.com is a new service which aims to help weed out phishing URLs and email addresses using wisdom of the crowds. Users can submit emails/URLs which they suspect of fraud and others can vote if they really are fraudulent or not. I think it is a great concept. There is a REST API using which applications can embed this webservice within them. So for example, there could be a outlook plugin which will display "phishy" email addresses in a special way in order to alert the user immediately. Same for web browsers which can render phishing websites in a special stylesheet. The applications can also add interface for the user to submit suspect pages and email easily without using web browsers.
I checked out the API and it does not feel like it is fully baked! There are interfaces for authorization and checking email/url status and submitting new emails/urls. Some things that stand out immediately are:
- Exclusive use of SSL for the API access.
- Parameter authentication (i.e. including cryptographic digest of all the parameters to ensure that parameters are not changed using man-in-the-middle attack)
- Choice of xml or php output.
The api calling sequence works like this:
- User registers on the web for API access and gets api key and shared secret
- Using the API, application gets a frob (what is behind the name ?) and authorization url using auth.frob.request
- User has to authorize the frob using the authorization url specified in the response. (optionally you can specify callback url which the server will call for authorization, I will need to check this from home when I have access to a server -- the docs are very thin about the mechanism)
- Once authorized, app uses the frob and gets a token for short time API access (30 minutes in my tests) (auth.token.request)
- App can check token status which tells remaining time on token.(auth.token.status)
- App can revoke the token when it is done using it. (auth.token.revoke)
- The APIs for check.url, check.email, submit.url, submit.email then use the token.
I did not understand why there is a need for FROB in this, why can't you just get the token from api key and shared secret ? What problem are they solving by this indirection ?
P.S. the check_url interface is not working, I am getting invalid token error. and the same token can be revoked successfully.
P.P.S. The API uses SSL (no cleartext api available) and ruby's open-uri library insists on checking the server SSL certificate which always fails (probably because signer needs to be trusted by openssl), I had to change it locally to ignore ssl verification in order to proceed.
Update (Oct/12/06): the check.url interface is finally working. For this API, the signature needs to be calculated before escaping the url. I refactored the ruby script a bit to remove redundant code and moved the configuration to a seperate file. I still need to work with the response parser and make it general for all types of responses. XML parsing gets so ugly so fast, it's amazing!