How we tested DLP products
- — 13 August, 2012 04:31
We tested multiple common endpoint scenarios, by trying to upload, print, email or otherwise transfer data that should be blocked, quarantined, warned about or simply monitored under the rules and policies we set up. The following specific types of endpoint tests were conducted:
- Email (various clients including Outlook and Mozilla Thunderbird)
- Web browser file upload (both Port 80 and Port 443), using multiple browsers.
- Copy to USB devices such as thumb drives and external hard drives
- FTP file uploads
- Print to printer and file
- Screen capture (PrintScreen, SnagIt, Windows Snipping Tool)
- Clipboard cut and paste
We utilized a variety of rules and policies, some built in to the product, some created with technical assistance from the vendor and some created by us in the lab. While our main focus was compliance with acceptable use and regulations such as PCI-DSS and HIPAA, we also set up custom rules to protect 'user-defined' confidential and classified documents.
If a test initially failed, we gave the vendor an opportunity to suggest rule changes or other workarounds. In most cases, tweaking DLP policies or using a different logic corrected the problem. This accommodation was made to compensate for each vendor's unique interpretation of how DLP rules should be set up and administered to achieve specific DLP objectives. For the actual endpoint testing, we did not utilize the many tools and methods available to us as network administrators to try to 'break' products, under the assumption that the typical end user would not have access to these tools. We did, however, test from low-privileged Windows accounts up to accounts with full administrator privileges.
Because of the virtually infinite combinations of network topologies, operating systems and hardware, we did not test for performance or scalability. These factors may only come into play in certain situations (large, geographically dispersed networks), and perhaps seldom, if ever, in others (smaller networks with less traffic).
Although some of the vendors (Websense and Cisco) provided us with appliances for the DLP host machine, we tested the endpoints on the lab's own equipment:
- Physical and virtual desktops running Windows 7 Professional
All the vendors offered to send engineers to set up their equipment and configure the test environment, but to maintain an independent test setting we chose to do our own installations, although we did consult most vendors for the initial install. For host DLP systems we utilized x86 or 64-bit Dell PowerEdge rack servers running either Windows Server 2008 R2 (for Sophos) or Linux (for Trend Micro). One of the DLP products, Verdasys' Digital Guardian, was tested as a hosted cloud solution.
Tests were conducted sequentially. In between testing of each product, we re-imaged both desktops and servers (except vendor-supplied hardware) and sequestered each product on a closed internal subnet with a single point (static IP address) to the Internet. At no time were any two products connected to the network at the same time, nor did client machines ever have more than one agent installed.
Read more about wide area network in Network World's Wide Area Network section.