A general fix needs some more thought. Looking at the possible fixes that were suggested, these only work in limited scenarios when you use relatively simple forms.
For example, they don't work in the following cases: - PK is not named id but something else, including composite PKs - all embedded forms as bind is never called on them One could think simply unsetting id entirely would do the trick. Unfortunately, some of the built-in validators rely on the id to determine whether an object is new, or already exists, e.g. the unique validators. What I do in these cases, is to simply not use the built- in validator, but a callback validator which has access to the underlying object directly. However, the unique validators could be changed so that you can pass an object to the validator, and it would not need the id for determining whether the object is new or not. Johannes On 10 Mai, 21:08, Stephen Melrose <[email protected]> wrote: > It's not that simple Russ. > > Where do you perform the check that they are allowed to access/save an > object of a certain ID? > > I personally have always checked the object after I've taken it from the > route and before I've passed it to the form. After I've done that, I don't > expect the object to magically be transformed into another record, a.k.a. > something they're not allowed to access, and I bet the vast majority of > symfony developers don't either. > > Your point is valid and one I've agreed with throughout this thread, if > you're restricting what a user can edit, you need to make sure you safe > guard your code properly, and that's something I didn't do purely because I > didn't expect the scenario I've detailed to occur, and to be honest, nor > should it. > > The ID is passed as a hidden field (for whatever reason), but I don't expect > the PK to be changed. How often would you actually edit a PK? > > I'm simply arguing the PK should be read only be default. > > On 10 May 2010 20:01, Russ <[email protected]> wrote: > > > > > Personally I always check if the user has credentials to edit the > > object anyway and I couldn't give a monkeys if they change the id > > using Firebug or whatever as long as it's to one they have access to. > > If not, they'll get a nice 403 response either way. > > > The way I see it, editing the ID using Firebug or some other method > > would be just the same as if they opened that object up for editing in > > the first place... As long as they are allowed to, then so be it. > > > On May 10, 12:16 pm, Stephen Melrose <[email protected]> wrote: > > > Hi, > > > > We have discovered what could be a potential flaw in the form > > > framework. The reason I'm discussing this here is because I'm in mixed > > > feelings as to whether this is actually bug or not, or rather poor > > > implementation on our part. Either way, I'm also saying this flaw > > > should be safe guarded against. > > > > We discovered that a malicious user can use the forms generated by the > > > form framework to edit content they shouldn't be able to. > > > > They do this by replacing the primary ID in the hidden form field with > > > that of the record they want to edit. When they hit save, the > > > validation is run, and the Object is updated with the new ID, so when > > > the save() is called, the other row is updated. > > > > Now, if we (as in developers) want to restrict editing of content for > > > certain users, then it is our responsibility to make sure we put safe > > > guards in place. I'm not arguing this fact. > > > > The reason I believe this to be a problem is how users will actually > > > guard their code. Most people (including myself) run all the safe > > > guard checks before the Object is passed into the Form on > > > construction. I don't then expect the POST data to override the > > > primary key of the Object on save. Infact, I can't think of an > > > instance I would ever want this to happen. > > > > I therefore propose that some sort of restriction/block is put in > > > place by default that stops the PK of an Object being altered on > > > bind(). > > > > Thoughts? > > > > Stephen Melrose > > > > -- > > > If you want to report a vulnerability issue on symfony, please send it to > > security at symfony-project.com > > > > You received this message because you are subscribed to the Google > > > Groups "symfony developers" group. > > > To post to this group, send email to [email protected] > > > To unsubscribe from this group, send email to > > > [email protected]<symfony-devs%[email protected]> > > > For more options, visit this group athttp:// > > groups.google.com/group/symfony-devs?hl=en > > > -- > > If you want to report a vulnerability issue on symfony, please send it to > > security at symfony-project.com > > > You received this message because you are subscribed to the Google > > Groups "symfony developers" group. > > To post to this group, send email to [email protected] > > To unsubscribe from this group, send email to > > [email protected]<symfony-devs%[email protected]> > > For more options, visit this group at > >http://groups.google.com/group/symfony-devs?hl=en > > -- > If you want to report a vulnerability issue on symfony, please send it to > security at symfony-project.com > > You received this message because you are subscribed to the Google > Groups "symfony developers" group. > To post to this group, send email to [email protected] > To unsubscribe from this group, send email to > [email protected] > For more options, visit this group > athttp://groups.google.com/group/symfony-devs?hl=en -- If you want to report a vulnerability issue on symfony, please send it to security at symfony-project.com You received this message because you are subscribed to the Google Groups "symfony developers" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/symfony-devs?hl=en
