This site is the archived OWASP Foundation Wiki and is no longer accepting Account Requests.
To view the new OWASP Foundation website, please visit

Difference between revisions of "Projects/OWASP Secure Application Design Project/Releases/Current"

Jump to: navigation, search
(Business Logic Invocation Technique)
(Blanked the page)
Line 1: Line 1:
This section describes some of the important design flaws that can leave a backdoor in the application to access it without authentication or manipulate its business logic. We will understand such flaws and secure design recommendations in detail.
==Data Binding Technique==
Another popular feature seen in most of the design frameworks today is data binding, where the request parameters get directly bound to the variables of the corresponding business/command object. Binding here means that the instance variables of such classes get automatically initialized with the request parameter values based on their names.
Consider a sample design given below; observe that the business logic class binds the business object with the request parameters.
'''What is the flaw?'''
The flaw in such design could be that the business objects may have variables that are not dependent on the request parameters. Such variables could be key variables like price, max limit, role etc. having static values or dependent on some server side processing logic.
A threat in such scenarios is that an attacker may supply additional parameters in request and try to bind values for unexposed variable of business object class.  As illustrated in the figure below, the attacker sends an additional “price” parameter in the request and binds with the unexposed variable “price” in business object, thereby manipulating business logic. 
'''Secure Design Recommendation:'''
*An important point to be noted here is that the business/form/command objects must have only those instance variables that are dependent on the user inputs. If additional variables are present those must not be vital ones like related to the business rule for the feature.
*In any case the application must accept only desired inputs from the user and the rest must be rejected or left unbound. And initialization of unexposed of variables, if any must take place after the binding logic. 
'''Review Criteria'''
Review the application design and check it is incorporates a data binding logic. In case it does, check if business objects/beans that get bound to the request parameters have unexposed variables that are meant to have static values. If such variables are initialized before the binding logic this attack will work successfully.
==Placement of Security Controls==
Placement of security checks is a vital area of review in an application design. Incorrect placements can render the applied security controls null and void. So, it is important to study the application design and spot the correctness of such checks in the overall execution flow of the design.
Most of the application designs are based on the concept of Model-View-Controller (MVC). They have a central controller, which listens to all incoming request and delegates control to appropriate form/business processing logic. And ultimately the user is rendered with a view. In such a layered design, when there are many entities involved in processing a request, developers often go wrong in placing the security controls at  the right place.
Most application developers feel “view” is the right place to have the security checks like authentication check etc.
'''What is the flaw?'''
It thus seems logical that if you restrict the users at the page/view level they won’t be able to perform any operation in the application. But what if instead of requesting for a page/view an unauthorized user tries to request for an internal action like to action to add/modify any data in the application? It will get processed but the resultant view will be denied to the user; because the flaw lies in just having a view based access control in the applications. I am sure you will agree that a lot of processing for a request is done before the “view” comes into picture in any design. So the request to process any action will get processed successfully without authorization.
Consider a MVC based given in the figure below. Observe in the figure that the authentication check is present only in the view pages.
Observe that neither the controller servlet (central processing entity) nor the action classes have any access control checks. So here, if the user requests for an internal action like add user details, etc. without authentication it will get processed, but the only difference is that the user will be shown an error page as resultant view will be disallowed to the user.
'''Insecure POST-BACK’s in ASP.NET'''
A similar flaw is predominantly observed in ASP.NET applications where the developers tend to mix the code for handling POSTBACK’s and authentication checks. Usually it is observed that the authentication check in the ASP.NET pages are not applied for POSTBACKs, as indicated below. Here, if an attacker tries to access the page without authentication an error page will be rendered. Instead, if the attacker tries to send an internal POSTBACK request directly without authentication it would succeed. A detailed explanation is present here -[[]]
'''Secure Design Recommendation:'''
Do not just restrict the access of the users only over views. It is imperative to place all validation checks before processing any business logic and in case of ASP.NET applications independent of the POSTBACKs.
'''Review criteria'''
Check if the placement of the security checks is correct. The security controls like authentication check must be place before processing any request.
==Execution Flow==
Execution flow is another important consideration of design. The execution flow must terminate appropriately in case of an error condition. However, due to mishandling of some programming entities there could be a big hole in the application which would allow unrestricted access to applications. One such flaw is related to – “sendRedirect” method in J2EE applications.
This method is used to send a redirection response to the user who then gets redirected to the desired web component, whose ULR is passed as an argument to the method –in this case home.html.
One such misconception is that execution flow in the Servlet or JSP page that is redirecting the user stops after a call to this  method.
Take a look at the code snippet below, it checks for authenticated session using an “if” condition.  If the condition fails the response.sendRedirect() is used to redirect the user to an error page.
Note that there is code present after the If condition, which continues to fetch request parameters and processes business logic for instance adding a new branch entry of a bank in this case.
'''What is the flaw?'''
This flaw manifests as a result of the misconception that the execution flow in the JSP/Servlet page stops after the “sendRedirect” call. However it does not; in this case the execution of the servlet would continue even if an invalid session is detected by the “if” condition and thus the business logic will get processed for unauthenticated requests.
'''Note:''' The fact that execution of a servlet or JSP continues even after sendRedirect() method, also applies to Forward method of the RequestDispatcher Class.
However, <jsp:forward> tag is an exception, it is observed that the execution flow stops after the use of <jsp:forward> tag.
'''Secure Design Recommendation:'''
Since this flaw results from the assumption made by developers that control flow execution terminates after a sendRedirect call, the recommendation would be to terminate the flow using a “return” statement.
'''Review criteria'''
Check if there is an appropriate logic to terminate the execution flow is present in case of an error condition. Check for similar instances of insecure security controls built using “sendRedirect” method.
Some links for reference:

Latest revision as of 18:29, 6 May 2014