Web application firewall is a simple term, but understanding what it means is proving so difficult for customers that an industry consortium is publishing advice on how to make a choice among the many devices that fall into this category.
On Monday, a 20-page document called Web Application Firewall Evaluation Criteria is being published by the Web Application Security Consortium, a group formed a year ago from vendors, consultants and end users. The document will be available here.
Broadly, Web application firewalls examine HTTP and HTTPS traffic at the application layer, looking for attacks masquerading as legitimate application traffic. They defend against attempts to tap sensitive information stored on Web application servers, such as credit card and social security numbers as well as proprietary corporate information.
There are such a wealth and variety of methods for accomplishing this goal that it is difficult for potential customers to figure out what product best suits their needs, says Mark Kraynak, director of product marketing for WAF vendor Imperva who served on the Web Application Security Consortium committee that wrote the document. Other vendors include Citrix, F5 Networks, NetContinuum and Protegrity among others.
"There's too much functionality and only one name for the products," says Kraynak. "A lot of our customers don't know what criteria they should be evaluating. There's confusion."
No single WAF device is appropriate for all networks, says Ivan Ristic, who headed the evaluation effort for the consortium. He also runs Thinking Stone, a Web application security consulting firm. "You need to look at your security requirements and business goals. Create a short list of features you need, and use that to sort products."
For example, a business that needs to document all HTTP transactions for regulatory purposes may need a WAF with very few features, Ristic says. Or a business with a single Web server might need no separate WAF, but only application firewall software that can run on the server itself, he says.
The range of features is broad. A WAF can deal with SSL traffic, for example, by terminating it, examining it and passing it on or capturing it and decrypting it but not terminating sessions. Similarly, if the WAF needs to block traffic, it can terminate network-protocol connections and not pass on malicious traffic or it can sever suspicious connections altogether.
Also, individual products can support more or fewer versions of HTTP, encryption and authentication. These devices may or may not support filtering of outgoing traffic as well.
"It's not like every Web application firewall needs all these features," Ristic says.
"Three years ago, there wasn't anything like this (evaluation criteria)," says a senior network executive at a mid-West financial services firm that uses Citrix gear. The firm would not allow its name to be used in this story. "We had to hire a third party to do attack and penetration testing for us."
Raphael San Miguel, who worked on the project and is also a senior consultant with daVinci Consulting, a security consulting firm in Spain, says management of these devices is always an important factor.
In particular, the devices should be able to adapt to new attacks without manual intervention. "You can't be changing configuration parameters all the time. If you do, security will depend on how good the administrator is who is working right now," he says.
The IT executive for the mid-West financial firm agreed. "It's not the type of solution you can buy, put in and never make changes again. It has to adapt to new threats," he says. Automation is important, he said, because the manual configuration of one of the devices he tested, made by Kavado, was so complex it was easy to misconfigure.
To do so, the devices must understand application logic and adapt their filters when they recognize traffic patterns that stray from how legitimate traffic performs, he says. Different devices determine normal traffic differently, he notes. Some have tools that simulate traffic to the server to figure out the give and take of legitimate traffic. Others monitor actual network traffic to make the same determination, he says. The former can discover quickly how an application works, but the latter is more accurate, he says.
San Miguel says that in a few weeks the consortium will publish applications that simulate attacks on applications that potential customers can use to test individual WAF products. Customers would set up a test network with a PC running the attack simulator and a server being attacked on which WAF devices can be installed to see how well they protect the server.
