rails: mass-assignment security concern with belongs_to relationships - ruby-on-rails

I've been reading up on rails security concerns and the one that makes me the most concerned is mass assignment. My application is making use of attr_accessible, however I'm not sure if I quite know what the best way to handle the exposed relationships is. Let's assume that we have a basic content creation/ownership website. A user can have create blog posts, and have one category associated with that blog post.
So I have three models:
user
post: belongs to a user and a category
category: belongs to user
I allow mass-assignment on the category_id, so the user could nil it out, change it to one of their categories, or through mass-assignment, I suppose they could change it to someone else's category. That is where I'm kind of unsure about what the best way to proceed would be.
The resources I have investigated (particularly railscast #178 and a resource that was provided from that railscast), both mention that the association should not be mass-assignable, which makes sense. I'm just not sure how else to allow the user to change what the category of the post would be in a railsy way.
Any ideas on how best to solve this? Am I looking at it the wrong way?
UPDATE: Hopefully clarifying my concern a bit more.
Let's say I'm in Post, do I need something like the following:
def create
#post = Post.new(params[:category])
#post.user_id = current_user.id
# CHECK HERE IF REQUESTED CATEGORY_ID IS OWNED BY USER
# continue on as normal here
end
That seems like a lot of work? I would need to check that on every controller in both the update and create action. Keep in mind that there is more than just one belongs_to relationship.

Your user can change it through an edit form of some kind, i presume.
Based on that, Mass Assignment is really for nefarious types who seek to mess with your app through things like curl. I call them curl kiddies.
All that to say, if you use attr_protected - (here you put the fields you Do Not want them to change) or the kid's favourite attr_accessible(the fields that are OK to change).
You'll hear arguments for both, but if you use attr_protected :user_id in your model, and then in your CategoryController#create action you can do something like
def create
#category = Category.new(params[:category])
#category.user_id = current_user.id
respond_to do |format|
....#continue on as normal here
end

OK, so searched around a bit, and finally came up with something workable for me. I like keeping logic out of the controllers where possible, so this solution is a model-based solution:
# Post.rb
validates_each :asset_category_id do |record, attr, value|
self.validates_associated_permission(record, attr, value)
end
# This can obviously be put in a base class/utility class of some sort.
def self.validates_associated_permission(record, attr, value)
return if value.blank?
class_string = attr.to_s.gsub(/_id$/, '')
klass = class_string.camelize.constantize
# Check here that the associated record is the users
# I'm leaving this part as pseudo code as everyone's auth code is
# unique.
if klass.find_by_id(value).can_write(current_user)
record.errors.add attr, 'cannot be found.'
end
end
I also found that rails 3.0 will have a better way to specify this instead of the 3 lines required for the ultra generic validates_each.
http://ryandaigle.com/articles/2009/8/11/what-s-new-in-edge-rails-independent-model-validators

Related

Check User Input Before Saving

Good Day All!
Edited for better understanding.
First model is Inventory and in this model I have Product_Type, Product_Name and User_ID.
Second model I have Users which consist of First_Name, Last_Name and Pin_Number.
On my Inventories page I have a form for checking out said Product_Type and Product_Name, also a place for a user to put their Pin_Number in. On submit, it will check the Pin_Number they have typed in and validate it in the Users model and if the Pin_Number is correct it will create an entry with said Product_Type, Product_Name and User_ID (which is pulled from Pin_Number that was submitted.)
I am just trying to figure out how to validate that Pin_Number they submitted.
Thats why I thought I had to do some kind of validation and an if statement based on that validation. Not sure how to go about that.
I hope this clears up any confusion.
I am just trying to figure out how to validate that Pin_Number they submitted.
What constitutes a valid pin_number? Just that it allows you to successfully look up a User? What if a user enters another user's pin_number? Is that considered 'valid'? Something to think about...
It would be helpful if you would add to your question what your params look like upon form submission. But, we can do some guess work.
So, let's assume that params looks something like:
{..., "inventory"=>{"product_type"=>"foo", "product_name"=>"Bar"}, "pin_number"=>5, ...}
In your controller, you'll probably do something like:
if #user = User.find_by(pin_number: params[:pin_number])
#inventory = Inventory.new(inventory_params)
#inventory.user = #user
if #inventory.valid?
#inventory.save
# perhaps do some other stuff...
else
# handle the case where the `#inventory` is not valid
end
else
# handle the case where the `#user` was not found
end
This assumes you have something like:
private
def inventory_params
params.require(:inventory).permit(:product_type, :product_name)
end
In your Inventory model, you probably want to do something like (I apologize, I'm not on Rails 5 yet, so some of the syntax may be incorrect):
class Inventory < ActiveRecord::Base
validates :user_id,
:product_type,
:product_name,
presence: true
belongs_to :user
end
You probably also want to consider adding an index on User.pin_number if you're going to be doing a lot of finding by it.
Not sure if I got the question right, but sounds like a custom validator to me.
You can learn more about custom validators in the official Rails documentation under "Custom Validators"
Also, consider moving the class for the custom validator you'll build to a concern, which is a great way to make it reusable. You can find more information on this StackOverflow question and this nice tutorial.

ActiveRecord uniqueness validation prevents update

I'm writing a web app using Rails, part of which includes giving users the ability to leave reviews for things. I wanted to put a validation in the review model to ensure that one user can't leave multiple reviews of the same item, so I wrote this:
class NoDuplicateReviewValidator < ActiveModel::Validator
def validate(record)
dup_reviews = Review.where({user_id: record.user,
work_id: record.work})
unless dup_reviews.length < 1
record.errors[:duplicate] << "No duplicate reviews!"
end
end
end
This validator has the desired behavior, i.e. it guarantees that a user can't review a work twice. However, it has the undesired side-effect that a user can't update an already existing review that he/she left. I'm using a really simple
def update
#review.update(review_params)
respond_with(#work)
end
in the reviews controller. How can I change either the validator or the update method so that duplicate reviews are prevented but updates are allowed?
I'm very new to Rails and web development, so I'm sure I've done something goofy here. I didn't use one of the built-in unique validators because what is unique is the user/work pair; there can more than one review by the same user of different works, and there can be more than one review of the same work by different users.
You can use validates_uniqueness_of on multiple attributes, like this:
validates_uniqueness_of :user_id, :scope => :work_id
Then a user would not be allowed to review a already reviewed work.
#Sharvy Ahmed's answer is definitely the best, as long as the case is simple enough – the OP's case seems like one of them.
However, if the conditions are more complex, you may need/want to write your custom validation. For that purpose, here's an example (checked with Rails 6.0).
class NoDuplicateReviewValidator < ActiveModel::Validator
def validate(record)
dup_reviews = Review.where(user_id: record.user,
work_id: record.work)
dup_reviews = dup_reviews.where.not(id: record.id) unless record.new_record?
if dup_reviews.count > 0
record.errors[:duplicate] << "No duplicate reviews!"
end
end
end
The idea is,
In create, all the relevant DB records retrieved with where can and should be used to judge the uniqueness. In the example new_record? is used to check it out, but it is actually redundant (because nil id matches no records).
In update, the DB row of the record to update must be excluded from the unique comparison. Otherwise, the update would always fail in the validation.
The count method is slightly more efficient in terms of DB transaction.

Rails forms - Should I build `accepts_nested_attributes_for` associations in the Controller, Model, or View?

The Question
I have a parent that accepts_nested_attributes_for a child. So, when I have a form for the parent, I need to build the child so I can display form fields for it as well. What I want to know is: where should I build the child? In the Model, View, or Controller?
Why I Am Asking This
You may be shaking your head and thinking I'm a madman for asking a question like this, but here's the line of thinking that got me here.
I have a Customer model that accepts_nested_attributes_for a billing_address, like so:
class Customer
belongs_to :billing_address, class_name: 'Address'
accepts_nested_attributes_for :billing_address
end
When I present a form for a new Customer to the user, I want to make sure there is a blank billing_address, so that the user actually sees fields for the billing_address. So I have something like this in my controller:
def new
#customer = Customer.new
#customer.build_billing_address
end
However, if the user doesn't fill out any of the billing_address fields, but tries to submit an invalid form, they will be presented with a form that no longer has fields for the billing_address, unless I put something like this in the create action of my controller:
def create
#customer = Customer.new(params[:customer])
#customer.build_billing_address if #customer.billing_address.nil?
end
There is another issue, which is that if a user tries to edit a Customer, but that Customer doesn't have an associated billing_address already, they won't see fields for the billing_address. So I have to add somethign like this to the controller:
def edit
#customer = Customer.find(params[:id])
#customer.build_billing_address if #customer.billing_address.nil?
end
And something similar needs to happen in the controller's update method.
Anyway, this is highly repetitive, so I thought about doing something in the model. My initial thinking was to add a callback to the model's after_initialize event, like so:
class CustomerModel
after_initialize :build_billing_address, if: 'billing_address.nil?'
end
But my spidey sense started tingling. Who's to say I won't instantiate a Customer in some other part of my code in the future and have this wreak havoc in some unexpected ways.
So my current thinking is that the best place to do this is in the form view itself, since what I'm trying to accomplish is to have a blank billing_address for the form and the form itself is the only place in the code where I know for sure that I'm about to show a form for the billing_address.
But, you know, I'm just some guy on the Internet. Where should I build_billing_address?
Though this advice by Xavier Shay is from 2011, he suggests putting it in the view, "since this is a view problem (do we display fields or not?)":
app/helpers/form_helper.rb:
module FormHelper
def setup_user(user)
user.address ||= Address.new
user
end
end
app/views/users/_form.html.erb:
<%= form_for setup_user(#user) do |f| %>
Note that I had to change the helper method to the following:
def setup_user(user)
user.addresses.build if user.addresses.empty?
user
end
The controller remains completely unchanged.
If you know your model should always have a billing address, you can override the getter for this attribute in your model class as described in the docs:
def billing_address
super || build_billing_address
end
Optionally pass in any attributes to build_billing_address as required by your particular needs.
You would use build if you want to build up something and save it later. I would say, use it in nested routes.
def create
#address = #customer.billing_addresses.build(params[:billing_address])
if #address.save
redirect_to #customer.billing_addresses
else
render "create"
end
end
Something like that. I also use the build when I'm in the console.
You have to remember the principles of MVC, which is to create DRY(don't repeat yourself) code, which is efficiently distributed between the various moving parts of the app
accepts_nested_attributes_for Is Great For Keeping Things DRY
accepts_nested_attributes_for is a model function which allows you to pass data through an association to another model. The reason why it exists is to give you the ability to populate another model's data based on a single form, and is excellent for extending functionality without too much extra code
The problem you're citing is that if you want to use the code in other areas of the app, you'll end up having all sorts of problems
My rebuttal to that is in order to create as efficient an application as possible, you want to write as little code as possible - letting Rails handle everything. The accepts_nested_attributes_for function does allow you to do this, but obviously has a cost, in that you have to accommodate it every time you want to use it
My recommendation is to use what you feel is the most efficient code you can, but also keep to conventions; as this will ensure speed & efficiency
You should handle all these scenarios in controller, since it is not a responsibility of model.
Just in terms of keeping things DRY, you can write a method,
def build_customer(customer)
customer.build_billing_address if customer.billing_address.nil?
#add more code if needed
end
And inside controller you can call this method wherever it is needed. e.g.
def create
#customer = Customer.new(params[:customer])
if #customer.save
redirect_to #customer.billing_addresses
else
build_customer(#customer)
render "new"
end
end

How to refactor that complex singelton model method for create nested models in Rails?

I have following complex method which I cut off from controller:
def self.create_with_company_and_employer(job_params)
company_attributes = job_params.delete(:company_attributes)
employer_attributes = job_params.delete(:employer_attributes)
new(job_params) do |job|
job.employer = Employer.find_or_create_by_email(employer_attributes)
company_attributes[:admin_id] = job.employer.id if Company.find_by_nip(company_attributes[:nip]).nil?
job.company = Company.find_or_create_by_nip(company_attributes)
Employment.create(employer_id: job.employer.id, company_id: job.company.id)
end
end
I using here two nested_attributes functionality for create company and employer.
Whole code you can find here: https://gist.github.com/2c3b52c35df763b6d9b4
company_attributes[:admin_id] = job.employer.id if Company.find_by_nip(company_attributes[:nip]).nil?
Employment.create(employer_id: job.employer.id, company_id: job.company.id)
Basically I would like to refactor that two lines:
I looked at your gist and i think this is a design issue.
your Employment and Job models seem somewhat redundant, but i don't know what are their actual purpose exactly so i can't really help for now on this matter (i have a hunch that your schema could be remodeled with the employements belonging to the jobs). However, if you really want to, you can use an after_create callback to manage the replication :
class Job < ActiveRecord::Base
after_create :create_corresponding_employment
def create_corresponding_employment
Employment.create( employer_id: employer.id, company_id: company.id )
end
end
this gets you rid of the last line of your method.
the other line you want to get rid of is tricky : you assign an admin_id to your company, but why would you want to do that ? In fact, you're just creating a 'hidden' relation between Company and Employer (a belongs_to one). Why do you need that ? Give more information and i can help.
one more thing: it is not advised to delete keys form the params, or even modify the hash directly. Use a copy.

Capturing current_user.id saving deeply nested records

Lets say I have a working form that looks like the following
=form_for #survey do |f|
=f.text_field :name
=f.fields_for :questions do |question_fields|
=question_fields.text_field :question_text
=question_fields.fields_for :answers do |answer_fields|
=answer_fields.text_field :answer_text
Because different parts of the form can be added and updated by different users I need a way to get the user_id into each model before it is saved. I realize it is not mvc compliant to be able to access current_user inside the model, that being said I am left without a solution.
If I was only saving one object it would be simple enough to assign the current_user.id to the object in the controller, but given the deeply nested nature of this form that starts to look like an ugly solution.
Is there an expert/railsy way to handle this?
Rails 3.2, devise
Can't each of the objects simply steal the user_id from their "parent" relationship? This is a common pattern:
class Answer < ActiveRecord::Base
before_validation :assign_user_id
protected
def assign_user_id
# Don't attempt if the question is not defined,
# or the user_id field is already populated.
return unless (self.question or self.user)
self.user_id = self.question.user_id
end
end
This involves a bit of additional database activity to resolve the answer for each question, as creating it in a scope is not sufficient, but it makes it pretty much fool-proof.
What you probably want to do is stuff in the user_id parameter when creating each record. This means your create call needs to merge in a :user_id key where required. The nested helper doesn't do this by default, though, so if you're using that you may just leave it up to the assign method.

Resources