Andres, >> Andres, when I have suggested this feature in w3af I didn't mean *full* REST >> specification support. >> >> Today a lot of web applications (especially based on frameworks like Django >> or in the old way by Apache mod_rewrite module) uses REST-like URLs e.g.: >> >> http://example.com/foo/bar/123 >> >> In this URL we (not scanner) can see such parts as: >> >> * foo - controller name >> * bar - action name >> * 123 - parameter value >> >> From classic web spider point of view it looks like directory hierarchy - it >> is incorrect behavior! All these parts we need to fuzz! > > Agreed, so we don't want to support REST, we want to support > mod_rewrite. Would that appreciation be correct? I wouldn't use term mod_rewrite as the name of the whole feature because mod_rewrite is simply one of REST URLs cases. In modern web applications which are based on frameworks like Django *internal URL processing* becomes more and more popular. mod_rewrite is web server based URL processing.
>> What I suggest to implement is rules for such URLs. It can be done as >> http-settings >> file option called "url-rules" (name is not important): >> >> /top/users/%s/view/%d/ >> /controller/action/%d/ >> ... >> >> %s and %d are special tokens which can be used by w3af to determine fuzz >> points. > > I like the idea, but I would do it in a more configurable way. > Right now we have the "fuzzFileName" setting in w3af, which > enabled/disables what I explained in the initial email. In the future > I would like to see the following options: > > * fuzzFileName (default: False) > * fuzzDirectories (default: False) > * url-rules (default: no filename with rules) > > With this, when a user wants to fuzz all the "directories" in all > URLs he just enables "fuzzDirectories" and "fuzzFileName". If he wants > to have more control over which parts of the URL are actually fuzzed, > he can disable the previous ones and set the url-rules himself. Hmmm, interesting idea. But let's call this option something like fuzzURLParts. fuzzDirectories can be misunderstood by user. > For the rules file, what I recommend is that we support parsing of > mod_rewrite and django rules (if possible) so that a developer can > simply copy/paste those rules into a file and point w3af into it. Agree, plus Nginx config > In your example you put something like %s and %d. Do we care if > it's a string or a digit? Should the scanner change it's behavior > based on that? I think it is not so important on the first iteration. We can consider all such tokens as strings. > > Regards, > >> >>> This email is just a conversation starter for defining how we're >>> going to deal with REST urls. >>> >>> REST, as described in [0], has two important moving parts: >>> 1- URLs that "look nice" (no parameters: /people/1/phones/23 ) >>> 2- Heavy usage of HTTP methods like GET, POST, DELETE, PUT. >>> >>> The first question that I would ask myself is... do we want to >>> support 1 and 2? Only 1? What is really needed by our users? >>> >>> If we only want to implement #1, it should be easy enough, since >>> we already have something similar (see: mutantFileName.py). This >>> mutant, together with the fuzzer.py (more specifically >>> _createFileNameMutants) will behave like this: >>> >>> - Original URL: http://host.tld/foo/spam-eggs.jsp >>> - Input strings: [ '<script>alert(1)</script>', 'ping localhost'] >>> - Output URLs: >>> * http://host.tld/foo/<script>alert(1)</script>-eggs.jsp >>> * http://host.tld/foo/spam-<script>alert(1)</script>.jsp >>> * http://host.tld/foo/ping%20localhost-eggs.jsp >>> * http://host.tld/foo/spam-ping%20localhost.jsp >>> >>> As you can see, it will split the filename using any character >>> that's not a letter and put the strings into those positions. If we >>> change this from just the filename into the whole path, it should work >>> and inject into each URL section. >>> >>> Please note that the current implementation only performs file >>> name fuzzing if misc-settings fuzzFileName is enabled (which is off by >>> default). Should we also think about this and potentially modify this >>> to true? >>> >>> Regarding #2 , I don't see a reason for it not to work with >>> w3af... but I could be mistaken. We should perform some tests to check >>> if w3af parses and correctly sends requests associated with forms that >>> use PUT, DELETE, etc. The meta-question here is... do we want w3af to >>> send requests that will "DELETE" stuff? >>> >>> Ok... that's enough for a conversation starter :) What do you guys >>> think? >>> >>> [0] http://microformats.org/wiki/rest/urls >>> >>> Regards, >> >> >> -- >> Taras >> http://oxdef.info >> > > > -- Taras http://oxdef.info ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity, and more. Splunk takes this data and makes sense of it. IT sense. And common sense. http://p.sf.net/sfu/splunk-novd2d _______________________________________________ W3af-develop mailing list W3af-develop@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/w3af-develop
