- Custom events - These frameworks implement non-standard or custom event binding mechanisms, and in the case for ReactJS, it creates a so-called “Virtual DOM” which provides an internal representation of events outside of the real browser DOM. It is important to discover the type and location of every actionable component on the page. Tracking the event bindings on a real DOM is relatively straightforward by shimming EventTarget.prototype.addEventListener and determining the event type, and which node it is bound to.
However, in cases where a framework manages its own event delegation (such as in the ReactJS Virtual DOM) it becomes more efficient to hook into the framework, effectively providing a query language into the framework for its events (instead of listening for them).
According to ReactJS page:
- React doesn't actually attach event handlers to the nodes themselves.
- When React starts up, it starts listening for all events at the top level using a single event listener.
- When a component is mounted or unmounted, the event handlers are simply added or removed from an internal mapping.
- When an event occurs, React knows how to dispatch it using this mapping.
AppSpider has now created a generalized lightweight framework hooking structure that can be used to effectively crawl/discover/scan frameworks that do things ‘their own way.' Look for an upcoming announcement on how you can incorporate and contribute your own custom framework scanning hooks with AppSpider.
So what is AppSpider doing with ReactJS now? AppSpider is leveraging Facebook's open source developer tools (react-devtools) that are wrapped in a generalized framework hook and now crawled exhaustively by AppSpider. Additionally, ‘doin' it their own way' event binding systems (such as the ReactJS Virtual DOM) are being considered and executed. It is still the case that frameworks are supported right out of the box including AngularJS, Backbone, jQuery, Knockout, etc., without the need for tuning. Only where needed are we adding specific support for frameworks with custom techniques.
Why is this important? Web application security scanners struggle with understanding these more complex types of systems that don't rely on fat clients and slow processes. Scanners were built using standard event names relying on these ever-present fields to allow them to interact with the web application. Without these fields, a traditional scanner no longer has the building blocks necessary to correctly interact with the web applications it is scanning. Additionally, they have used the ever-present DOM structure to better understand the application and assist with crawling. This becomes difficult if not impossible for a traditional scanner when they have to deal with applications that process information on the server side instead of the client side. If this creates such an issue, why are people flocking to these frameworks? There are several reasons:
- Two-way data bindings which allow a better back and forth between client side and server side interactions
- Processing information on the server side which increases performance and gives a better user experience
- Flexibility to break away from the fat client side frameworks
These capabilities can make a dramatic difference to developers and end users but they also introduce unique issues to security teams.
Security teams and their tools are all used to the standard event names like OnClick or OnSubmit. These events drive how we interact with a web application, allowing our standard tools to crawl through the application and interact with them. By using these standard events we have been able to automate the manual tasks of making the application think we interacted with it. This becomes much more complicated when we introduce custom event names. How do you automate interacting with something that changes from application to application or even worse whenever you refresh the same application? AppSpider answers that question by allowing you to connect directly into the framework and have the framework tell you what those custom events are before it even begins to crawl/attack.
Security experts have relied upon the DOM to know what was needed to test an application and monitor this interaction to understand potential weaknesses. Server side processing complicates this as all processing is done on the server side, away from the eyes and tools of the security expert and displaying only the end results. With AppSpider, you can now handle applications that are utilizing server side processes because we are not dependent on what is shown to us; instead we already know what is there.
Currently, the only way for pen testers to conduct web application tests on applications using ReactJS and other modern formats is to manually attack them, working their way one-by-one through each option. This is a time consuming and tedious task. Pen testers lack the tools to be able to quickly and dynamically scan a web application using these SPA frameworks to identify potential attack points and narrow down where they would like to do further manual testing. AppSpider now allows them to quickly and efficiently scan these applications, saving them time and allowing them to focus efforts where they will be the most effective.
How can you tell if your scanner is supporting these custom event names? Answering this question can be difficult as you have to know your application to truly understand what is being missed. You can typically see this quickly when you start to analyze the results of your scans. You will see areas of your application completely missed or parameters that don't show up in your logs as being tested against.
Know your weaknesses. Can your scanner effectively scan this basic ReactJS application and find all of the events? http://webscantest.com/react/ Test it and find out! Is your scanner able to get see past the DOM? AppSpider can.