Just when I thought I was done talking about DLP, interest starts to increase again. Below is an article I wrote up on how to minimize the complexity of a DLP deployment. This was for the Websense customer newsletter/site, but is my usual independent perspective.
One of the most common obstacles to a DLP deployment is psychological, not technical. With massive amounts of content and data streaming throughout the enterprise in support of countless business processes, the idea that we can somehow wrangle this information in any meaningful way, with minimal disruptions to business process, is daunting if not nigh on inconceivable. This idea is especially reinforced among security professionals still smarting from the pain of deploying and managing the constant false positives and tuning requirements of intrusion detection systems.
Since I started to cover DLP technologies about 7 years or so ago, I’ve talked with hundreds of people who have evaluated and/or deployed data loss prevention. Over the course of those conversations I’ve learned what tends to work, what doesn’t, and how to reduce the potential complexity of DLP deployments. Once you break the process down it turns out that DLP isn’t nearly as difficult to manage as some other security technologies, and even very large organizations are able to rapidly reap the benefits of DLP without creating time-consuming management nightmares.
The trick, as you’ll see, is to treat your DLP deployment as an incremental process. It’s like eating an elephant – you merely have to take it one bite at a time. Here are my top 3 tips, drawn from those hundreds of conversations:
1. Narrow your scope:
One of the most common problems with an initial DLP deployment is trying to start on too wide a scale. Your scope of deployment is defined by two primary factors – how many DLP components you deploy, and how many systems/employees you monitor. A full-suite DLP solution is capable of monitoring network traffic, integrating with email, scanning stored data, and monitoring endpoints. When looking at your initial scope, only pick one of the components to start with.
I usually recommend starting with anything other than endpoints, since you then have fewer components to manage. Most organizations tend to start on the network (usually with email) since it’s easy to deploy in a passive mode, but I do see some companies now starting with scanning stored data due to regulatory requirements.
In either case, stick with one component as you develop your initial policies and then narrow the scope to a subset of your network or storage. If you are in a mid-sized organization you might not need to narrow too much, but in large organizations you should pick a subnet or single egress point rather than thinking you have to watch everything.
Why narrow the scope? Because in our next step we’re going to deploy our policies, and starting with a single component and a limited subset of all your traffic/systems provides the information you need to tune policies without being overwhelmed with incidents you feel compelled to manage.
2. Start with one policy:
Once you’ve defined your initial scope, it’s time to deploy a policy. And yes, I mean a policy, not many policies. The policy should be narrow and align with your data protection priorities; e.g. credit card number detection or a subset of sensitive engineering plans for partial document matching.
You aren’t trying to define a perfect policy out of the box; that’s why we are keeping our scope narrow. Once you have the policy ready, go ahead and launch it in monitoring mode. Over the course of the next few days you should get a good sense of how well the policy works and how you need to tune it. Many of you are likely looking for similar kinds of information, like credit card numbers, in which case the out of the box policies included in your DLP product may be sufficient with little to no tuning.
3. Take the next bite:
Once you are comfortable with the results you are seeing it’s time to expand your deployment scope. Most successful organizations start by expanding the scope of coverage (systems scanned or network traffic), and then add DLP components to the policy (storage, endpoint, other network channels).
Then it’s time to start the process over with the next policy.
This iterative approach doesn’t necessarily take very long, especially if you leverage out of the box policies. Unlike something like IDS you gain immediate benefits even without having to cover all traffic throughout your entire organization. You get to tune your policies without being overwhelmed, while managing real incidents or exposure risks.