Research

It seems blog popularity is a double edged sword. Yes, thousands of folks read our stuff every day. But that also means we are a target for many SEO Experts, who want to buy links from us. No, we don’t sell advertising on the site. But that doesn’t stop them from pummeling us with a bunch of requests each week. Most of the time we are pretty cordial, but not always. Which brings us to today’s story. It seems Rich was a little uppity yesterday and decided to respond to the link request with a serious dose of snark. Rich: Our fee is $10M US. Cash. Non-sequential bills which must be hand delivered on a unicorn. And not one of those glued-on horn jobs. Must be the real thing with a documented pedigree. I guess Rich thought that it was yet another bot sending a blind request and that his list of demands would disappear into the Intertubes, but alas, it wasn’t a bot at all. This SEO fellow and Rich then proceeded to debate the finer issues of unicorn delivery. Interestingly enough, the $10MM fee didn’t seem to be an issue. SEO Guy: Thanks for getting back. I may have some issues fulfilling your request. The $10M will not be a problem, however I don’t know if you’ve noticed, but unicorns are a heavily-endangered species. Even to rent one would require resources that exceed my nearly limitless budget. Do you know how much a unicorn pilot charges by the hour? Rich: African or European unicorn? SEO Guy: How far do you live from Ireland? Rich: About 7000 miles, but my wife has unknown ancestors still living there and I have red hair. Not sure if that will get a discount. SEO Guy: Would it be okay if the unicorn itself delivered the (what I am assuming is a golden satchel of) money instead? I know you want it hand-delivered (mind out of the gutter) and that unicorns lack hands. Rich: Excellent point and I see that will save on the piloting fees. Yes, but only if we can time delivery for my daughter’s birthday and you also include a frosted cupcake with a candle on it for her. I think she’d like that. You can deduct the cost of the cupcake from the $10M, if that helps…but not the cost of the candle. So yes, as busy as we are with launching our super sekret project, polishing the CCSK training course, and all our client work, we still have time to give a hard time to a poor sap trying to buy a few links for his SEO clients. So every time I’m grumpy because QuickBooks Online is down, the EVDO service in my favorite coffee shop is crap, and I have to restructure a white paper – I can just appreciate the fact that I’m not the SEO guy. Yes, I do have to deal with asshats every day. But they are asshats of my own choosing. This guy doesn’t get to choose who he solicits and I’m sure a debate about unicorns was the highlight of his day of drudgery. Yes, I’m a lucky guy, and sometimes I need an SEO unicorn to remind me. -Mike Photo credits: “Unicorns!” originally uploaded by heathervescent Incite 4 U Testing my own confirmation bias: There are many very big-brained folks in security. Errata’s Rob Graham is one of them. Entering a debate with Rob is kind of like fighting a lion. You know you don’t have much of a chance; you can only hope Rob gets bored with you before he mauls your arguments with well-reasoned responses. So when Rob weighed in on Risk Management and Fukushima, I was excited because Rob put into words many of the points I’ve been trying (unsuccessfully) to make for years about risk management. But to be clear, I want to believe Rob’s arguments, because I am no fan of risk metrics (at least the way we practice them today). His ideas on who is an expert (and how that changes), and what that expert needs to do (have the most comprehensive knowledge of all the uncertainties) really resonated with me. Maybe you can model it out, maybe you can’t. But ultimately we are playing the odds and that’s a hard thing to do, which is why we focus so heavily on response. Now Alex Hutton doesn’t back down and has a well reasoned response as well. Though it seems (for a change) that both Rob and Alex are talking past each other. Yes, my appreciation of Rob’s arguments could be my own biases (and limited brainpower) talking, which wouldn’t be the first time. – MR Careful with that poison: Some days the security industry is like cross-breeding NASCAR with one of those crappy fashion/cooking/whatever reality shows. Everyone’s waiting for the crash, and when it happens they are more than happy to tell you how they would have done it better. As analysts we get used to the poison pill marketing briefs. You know, the phishing email or press release designed to knock the competition down. And there is no shortage of them filling my inbox after the RSA breach. At least NASCAR has the yellow caution flag to slow things down until they can get the mangled cars off the track. But I have yet to see one brief that shows any understanding of what happened or customer risk/needs. So I either delete them without reading or send back a scathing response. I have yet to see one of these work with a customer/prospect, so it all comes off as little more than jealous sniping. And besides, I know RSA isn’t the first security company to be breached, just one of the first to disclose, and I doubt any of the folks sending out this poison could survive the same sort of attack. If they aren’t already pwned, that is. (No link for this one since you all are probably getting the same emails). – RM No poop in the sandbox: Good article in Macworld describing the

I did not see the original Agile Ruined My Life post until I read Paul Krill’s An agile pioneer versus an ‘agile ruined my life’ critic response today. I wish I had, as I would have used Mr. Markham’s post as an example of the wrong way to look at Agile development in my OWASP and RSA presentations. Mr. Markham raises some very good points, but in general the post pissed me off: it reeks of irresponsibility and unwillingness to own up to failure. But rather than go off on a tirade covering the 20 reasons that post exhibits a lack of critical thinking, I’ll take the high road. Jon Kern’s quotes in the response hit the nail on the head, but did not include an adequate explanation of why, so I offer a couple examples. I make two points in my Agile development presentation which are relevant here. First: The scrum is not the same thing as Agile. Scrum is just a technique used to foster face-to-face communication. I like scrum and have had good success with it because, a) it promotes a subtle form of peer pressure in the group, and b) developers often come up with ingenious solutions when discussing problems in an open forum. Sure, it embodies Agile’s quest for simplicity and efficiency, but that’s just facility – not the benefit. Scrum is just a technique, and some Agile techniques work in particular circumstances, while others don’t. For example, I have never got pair programming to work. That could be due to the way I paired people up, or the difficulty of those projects might have made pairs impractical, or perhaps the developers were just lazy (which definitely does happen). The second point is that people break process. Mr. Markham does not accept that, but sorry, there are just not that many variables in play here. We use process to foster and encourage good behavior, to minimize poor behaviors, and to focus people on the task at hand. That does not mean process always wins. People are brilliant at avoiding responsibility and disrupting events. I couch Agile pitfalls in terms of SDL – because I am more interested in promoting secure code development – but the issues I raise cause general project failures as well. Zealots. Morons. Egoists. Unwitting newbies. People paranoid about losing their jobs. All these personality types figure into the success (or lack thereof) of Agile teams. Sometimes Agile looses to that passive-aggressive bastard at the back of the room. Maybe you need process adjustments, or perhaps better process management, or just maybe you need better people. If you use a hammer to drive a screw into the wall, don’t be surprised when things go wrong. Use the wrong tool or technique to solve a problem, and you should expect bad things to happen. Agile techniques are geared toward reducing complexity and improving communication; improvements in those two areas mean better likelihood of success, but there’s no guarantee. Especially when communication and complexity are not your problem. Don’t blame the technique – or the process in general – if you don’t have the people to support it. Share:

Ben Franklin was a pretty smart dude. My favorite quote of his is: “In this world nothing is certain but death and taxes.” For a couple hundred years, that was pretty good. But at this point, I’ll add mergers and acquisitions as the third certainty in this world. Maybe also that your NCAA bracket will get busted by some college you’ve never heard of (WTF VCU?). We saw this over the weekend. AT&T figures it’s easier and cheaper to drop $39 billion buying T-mobile than build their own network (great analysis by GigaOm) or gain market share one customer at a time. And in security, there are always plenty of deals happening or about to happen. Remember, security isn’t a standalone market over time, so pretty much all security companies will be folded into something or other. Take, for instance, WebSense trying to sell for $1 billion. And no, I’m not going to comment on whether WBSN is worth a billion. That’s another story for another day. Or the fact that given Intel’s balance sheet, McAfee will likely start taking down bigger targets. All we can count on is that there will be more M&A. But let’s take a look at why deals tend to be the path of least resistance for most companies. Outsourced R&D: Anyone who’s ever worked in a large company knows how hard it is to innovate internally. There is a lot of inertia and politics to overcome to get anything done. In many cases it’s easier to just buying some interesting technology, since the buyer has a snowball’s chance in hell of building it in-house. Distribution leverage: There are clear economies of scale in most businesses. So the more stuff in a rep’s bag and the bigger their market share, the more likely they’ll be able to sell something to someone. That’s what’s driving Big IT to continue buying everything. This also drive deals like AT&T/T-Mobile, because they are buying not just the network, but also the customers. Two drunks holding each other up: Yep, we also see deals involving two struggling companies, basically throwing a hail mary pass in hopes of surviving. That doesn’t usually work out too well. And those are just off the top of my head. I’m sure there are another 5-10 reasonable justifications, but from an end-user standpoint let’s cover some of the planning you have to make for the inevitable M&As. We will break the world up into BD (before deal) and AD (after deal). Before Deal: Assess vendor viability: First assess all your security vendors. Rank them on a scale from low viability (likely to be acquired or go out of business) to rock solid. Assess product criticality: Next look at all your security products and rate them on a scale from non-critical to “life is over if it goes down.” Group into quadrants: Using vendor viability and product criticality, you can group all your products into a few buckets. I recommend 4 because it’s easy. This chart should give you a good feel for what I’m talking about. Define contingency plans: For products in the “Get Plan B now” bucket, make sure you have clear contingency plans. For the other quadrants, think about what you’d do if there was M&A activity for those offerings, but they are less urgent than having a plan for the critical & fragile items. After Deal: Call your rep: Odds are your rep will be a pretty busy guy/gal in the days after a deal is announced. And there is a high likelihood they won’t know any more than you. But get in line and hear the corporate line about how nothing will change. Yada yada yada. Then, depending how much leverage you have, ask for a meeting with the buyer’s account team. And then extract either some pricing or product concessions. The first renewal right after a deal closes is the best time to act. They want to keep you (or the deal looks like crap), so squeeze and squeeze hard. Call the competition: Yes, the competition will be very interested in getting back in, hoping they can use the deal’s uncertainty as a wedge. Whether you are open to swapping out the vendor or not, bring the other guys in to provide additional leverage. Revisit contingency plans: You might have to pull the trigger even if you don’t want to, so it’s time to take the theoretical plan you defined before the deal, and adjust it for reality now that the deal has occurred. Evaluate what it would take to switch, assess the potential disruption, and get a very clear feel for how tough it would be to move. Don’t to share that information with vendors, but you need it. None of this stuff is novel, but it’s usually a good reminder of the things you should do, but may not get around to. Given the number of deals we have seen already this year, and the inevitably accelerating deal flow, it’s better to be safe than sorry. Share:

Now that we understand the technical architecture, let’s look at the principal features seen across most File Activity Monitoring tools. Entitlement (Permission/Rights) Analysis and Management One of the most important features in most FAM products is entitlement (permission) analysis. The tool collects all the file and directory permissions for the repository, ties them back to users and groups via directory integration, and generates a variety of reports. Knowing that an IP address tried to access a file might be somewhat useful but practical usefulness requires that policies be able to account for users, roles, and their mappings to real-world contexts such as business units. As we mentioned in the technical architecture section; all FAM products integrate with directory servers to gather user, group, and role information. This is the only way tools can gather sufficient context to support security requirements, such as tracing activity back to a real employee rather than just a username that might not indicate the person behind it. (Not that FAM is magic – if your directories don’t contain sufficient information for these mappings you still might have a lot of work to trace back identities). At the most basic level a FAM tool uses this integration to perform at least some minimal analysis on users and groups. The most common is permission analysis – providing complete reports on which users and groups have rights to which directories/repositories/files. This is often a primary driver for buying the FAM tool in the first place, as such reports are often required for compliance. Some tools include more advanced analysis to identify entitlement issues – especially rights conflicts. For example, you may be able to identify which users in accounting also have engineering rights. Or list users with multiple roles that violate conflict of interest policies. While useful for security, these capabilities can be crucial for finding and fixing compliance issues. A typical rights analysis will collect existing rights, map them to users and groups, help identify excessive permissions, and identify unneeded rights. Some examples are: Determine which users outside engineering have rights to engineering documents. Find which users with access to healthcare records also have access to change privileges, but aren’t in an administrative group. Identify all files and repositories the accounting group has access to, and then which other groups also have access to those files. Identify dormant users in the directory who still have access to files. Finally, the tool may allow you to manage permissions internally so you don’t have to manually connect to servers in order to make entitlement changes. Secure Aggregation and Correlation As useful as FAM is for a single repository, its real power becomes clear as you monitor larger swaths of your organization and can centrally manage permissions, activities, and policies. FAM tools use a similar architecture to Database Activity Monitoring – with multiple sensors, of different types, sending data back to the central management server. This information is normalized, stored in a secure repository, and available for a variety of analyses and reports. As a real-time tool the information is also analyzed for policy violations and (possible) enforcement actions, which we will discuss later. The tools don’t care if one server is a NAS, another a Windows server, and the last a supported document management system – it’s capable of reviewing all their contents consistently. This aggregation also supports correlation – meaning you can build policies based on activities occurring across different repositories and users. For example, you can alert on unusual activity by a single user across multiple file servers, or on multiple user accounts all accessing a single file in one location. Essentially, the FAM tool gives you a big picture view of all file activity across monitored repositories, with various ways of building alerts and analyzing the data, from a central management server. If your product supports multiple file protocols, it will present this in a consistent, activity-based format (e.g., open, delete, privilege change, etc.). Activity Analysis While understanding permissions and collecting activity are great, and may be all you need for a compliance project, the real power of FAM is its capability to monitor all file activity (at the repository level) in real time, and generate alerts, or block activity, based on security policies. Going back to our technical architecture: activity is collected via network monitoring, software agent, or other application integration. The management server then analyzes this activity for policy violations/warnings such as: A user accessing a repository they have access to, but have not accessed within the past 180 days. A sales employee downloading more than 5 customer files in a single day. Any administrator account accessing files in a sensitive repository. A new user (or group) being given rights to a sensitive directory. Any user account copying an entire directory from an engineering server. A service account accessing files. Some tools allow you to define policies based on a sensitivity tag for the repository and user groups (or business units), instead of having to manually build policies on a per-repository or per-directory level. This analysis doesn’t necessarily need to happen in real time – it can also be done on a scheduled or ad hoc basis to support a specific requirement, such as an auditor who wants to know who accessed a file, or as part of an incident investigation. We’ll talk more about reporting later. Data Owner Identification Although every file has an ‘owner’, translating that to an actual person is often a herculean process. Another primary driver of File Activity Monitoring is to help organizations identify file owners. This is typically done through a combination of privilege and activity analysis. Privileges might reveal a file owner, but activity may be more useful. You could build a report showing the users who most often access a file, then correlate that to who also has ownership permissions, and the odds are they will help quickly identify the file owner. This is, of course, much simpler if the tool was already monitoring a repository and can identify who initially created the file.

As this is posting, RSA is releasing a new SecureCare note and FAQ for their clients (Login required). This provides more specific prioritized information on what mitigations they recommend SecurID clients take. To be honest they really should just come clean at this point. With the level of detail in the support documents it’s fairly obvious what’s going on. These notes are equivalent to saying, “we can’t tell you it’s an elephant, but we can confirm that it is large, grey, and capable of crushing your skull if you lay down in front of it. Oh yeah, and it has a trunk and hates mice.” So let’s update what we know, what we don’t, what you should do, and the open questions from our first post: What we know Based on the updated information… not much we didn’t before. But I believe RSA understands the strict definition of APT and isn’t using the term to indicate a random, sophisticated attack. So we can infer who the actor is – China – but RSA isn’t saying and we don’t have confirmation. In terms of what was lost, the answer is, “an elephant” even if they don’t want to say so. This means either customer token records or something similar, and I can’t think of what else it could be. Here’s a quote from them that makes it almost obvious: To compromise any RSA SecurID deployment, the attacker needs to possess multiple pieces of information about the token, the customer, the individual users and their PINs. Some of this information is never held by RSA and is controlled only by the customer. In order to mount a successful attack, someone would need to have possession of all this information. If it were a compromise of the authentication server software itself, that statement wouldn’t be accurate. Also, one of their top recommendations is to use long, complex PINs. They wouldn’t say that if the server was compromised, which means it pretty much has to be related to customer token records. This also leads us to understand the nature of a potential attack. The attacker would need to know the username, password/PIN, and probably the individual assigned token. Plus they need some time and luck. While extremely serious for high-value targets, this does limit potential exposure. This also explains their recommendations on social engineering, hardening the authentication server, setting PIN lockouts, and checking logs for ongoing bad token/authentication requests. I think his name is Babar. What we don’t know We don’t have any confirmation of anything at this point, which is frankly silly unless we are missing some major piece of the puzzle. Until then it’s reasonable to assume a single sophisticated attacker (with a very tasty national cuisine), and compromise of token seeds/records. This reduces the target pool and means most people should be in good shape with the practices we previously recommended (updated below). One big unknown is when this happened. That’s important, especially for high-value targets, as it could mean they have been under attack for a while, and opponents might have harvested some credentials via social engineering or other means already. We also don’t know why RSA isn’t simply telling us what they lost. With all these recommendations it’s clear that the attacker still needs to be sophisticated to pull off more attacks with the SecurID data, and needs to have that data, which means customer risk is unlikely to increase if they reveal more. This isn’t like a 0-day vulnerability, where merely knowing it’s out there is a path to exploitation. More information now will only reduce customer risk. What you need to do Here are our updated recommendations: Remember that SecurID is the second factor in a two-factor system… you aren’t stripped naked (unless you’re going through airport security). Assuming it’s completely useless now, here is what you can do: Don’t panic. Although we don’t know a lot more, we have a strong sense of the attacker and the vulnerability. Most of you aren’t at risk if you follow RSA’s recommendations. Many of you aren’t on the target list at all. Talk to your RSA representative and pressure them for increased disclosure. Read the RSA SecureCare documentation. Among other things, it provides the specific things to look for in your logs. Let your users with SecurIDs know something is up and not to reveal any information about their tokens. Assume SecureID is no longer effective. Review passwords/PINs tied to SecurID accounts and make sure they are strong (if possible). If you change settings to use long PINs, you need to get an update script from RSA (depending on your product version) so the update pushes out properly. If you are a high-value target, force a password change for any accounts with privileges that could be seriously damaging (e.g., admins). Consider disabling accounts that don’t use a password or PIN. Set authentication attempt lockouts (3 tries to lock an account, or similar). The biggest changes are a little more detail on what to look for, which supports our previous assumptions. That and my belief their use of the term APT is accurate. Open questions I will add in my own answers where we have them: While we don’t need all the details, we do need to know something about the attacker to evaluate our risk. Can you (RSA) reveal more details? Not answered, but reading between the lines this looks like true APT. How is SecurID affected and will you be making mitigations public? Partially answered. More specific mitigations are now published, but we still don’t have full information. Are all customers affected or only certain product versions and/or configurations? Answered – see the SecureCare documentation, but it seems to be all current versions. What is the potential vector of attack? Unknown, so we are still assuming it’s lost token records/seeds, which means the attacker needs to gather other information to successfully make an improper authentication request. Will you, after any investigation is complete, release details so the rest of us can learn from your victimization? Answered. An RSA contact told me they have every

It’s hard to believe, but we have wrapped up the initial research on this series dealing with how network security evolves, given the need to provide access to critical information at any time, from anywhere, on any device. We call it any computing. We’ve dealt with the risks and how enforcement and policies will change. And talked quite a bit about integrating these enforcement points into the existing network and security infrastructure. Finally, we wrapped the series yesterday with Quick Wins, about the process of selecting and implementing these technologies. So here is the index of posts. Enjoy. The Risks Containing Access Enforcement Policy Granularity Integration Quick Wins If you missed any of these posts, check out our Complete Feed on the web or via RSS. Then you’ll be sure to get everything we publish. The next step is to assemble these posts, massage a bit, have someone who knows how to write edit the whole thing, and then publish as a white paper. That should happen over the next two weeks. Stay tuned – we’ll post the paper’s availability right here. Share:

We have gotten a bunch of questions about what people should do, so I thought I would expand more on the advice in our last post, linked below. Since we don’t know for sure who compromised RSA, nor exactly what was taken, nor how it could be used, we can’t make an informed risk decision. If you are in a high-security/highly-targeted industry you probably need to make changes right away. If not, some basic precautions are your best bet. Remember that SecurID is the second factor in a two-factor system… you aren’t stripped naked (unless you’re going through airport security). Assuming it’s completely useless now, here is what you can do: Don’t panic. We know almost nothing at this point, and thus all we can do is speculate. Until we know the attacker, what was lost, how SecurID was compromised (assuming it was), and the potential attack vector we can’t make an informed risk assessment. Talk to your RSA representative and pressure them for this information. Assume SecureID is no longer effective. Review passwords tied to SecurID accounts and make sure they are strong (if possible). If you are a high-value target, force a password change for any accounts with privileges that could be overly harmful (e.g., admins). Consider disabling accounts that don’t use a password or PIN. Set password attempt lockouts (3 tries to lock an account, or similar). I hope we’re wrong, but that’s the safe bet until we hear more. And remember, it isn’t like Skynet is out there compromising every SecurID-‘protected’ account in the world. Share:

One of the more interesting results from the Pwn2Own contest at CanSecWest was the exploitation of a Blackberry using a WebKit vulnerability. RIM just learned a lesson that Apple (and others) have been struggling with for a few years now. While I don’t think open code is inherently more or less secure than proprietary code, any time you include external code in your platform you are intrinsically tied to whoever maintains that code. This is bad enough for applications and plugins like Adobe Flash and Acrobat/Reader, but it is really darn ugly for something like Java (a total mess from a security standpoint). While I don’t know if it was involved in this particular hack, one of the bigger problems with using external code is when a vulnerability is discovered and released (or even patched) before you include the patch in your own distribution. Many of the other issues around external code are easier to manage, but Apple clearly illustrates what appears to be the worst one. This is the delay between initial release of patches for open projects (including WebKit, driven by Apple) and their own patches – often months later. During this window, the open source repository shows exactly what changed and thus points directly at their own vulnerability. As Apple has shown – even with WebKit, which it drives – this is a serious problem and seriously aggravates the wait for patch delivery. At this point I should probably make clear that I don’t think including external code (even open source) is bad – merely that it brings this pesky security issue which requires management. There are three ways to minimize this risk: Patch early and often. Keep the window of vulnerability for your platform/application as short as possible by burning the midnight oil once a fix is public. Engage deeply with the open source community your code comes from. Preferably have some of your people on the core team, which only happens if they actually contribute something of significance to the project. Then prepare to release your patch at the same time the primary update is released (don’t patch before – that might well break trust). Invest in anti-exploitation technologies that hopefully mitigate any vulnerabilities, no matter the origin. The real answer is you need to do all three. Issue timely fixes when you get caught unaware, engage deeply with the community you now rely on, and harden your platform. Share:

We have worked quickly through the main concepts of using network security tactics to provide access to the myriad of endpoint and mobile devices, so now let’s shift to a process to ensure success for your project. This is all about success, so we find the best path is to focus your project on establishing an initial quick win, and then gradually build momentum for the technology with expanded deployment. Step 1: Define Success We know this seems obvious, but it’s amazing how many organizations just start projects without focusing on the problem to solve and how to gauge success. So we start every process by making sure everyone is on the same page regarding what needs to be protected, and from what specific threats. You can do a formal threat model or an informal list of use cases. But you need to know, and everyone else must agree, what success means for this project. Step 2: Establish Deployment Plan What’s next? Protect the most critical information, of course. In this step get everyone on the same page regarding where enforcement points will be installed and how you’ll phase in the deployment. Understand up front that you will be wrong – what makes the most sense may change as you go through the project. This isn’t about carving anything in stone – it’s thinking ahead of time about the best way to solve your problem – before some vendor puts you on a runaway train. Note that all this work happens before you start engaging with vendors. We advocate a strong plan before starting product evaluation. Again, things may change, but if you don’t know what you are trying to get done ahead of time, the odds are you will never get there. Step 3: Technology Evaluation Now you get to suffer though any number of dog and pony shows to establish your short list of vendors. We suggest keeping the meetings focused and making sure you do some homework before sitting with a vendor. Then you’ll at least know when they are blatantly pulling your leg. Step 4: PoC When dealing with complicated technology, we always recommend a proof of concept (PoC) before buying anything. Given the number of integration points for Network Access Control, you’d be crazy not to ensure each vendor could work with your existing stuff. We also believe the PoC needs to be customer driven; which means you define the use cases, integration points, and management tasks to be tested – not the vendor. Surprisingly enough, vendors have a unfortunate tendency to direct you toward the strengths of their products. You need to stay laser focused on solving your problem. Be particularly wary of user experience and day-to-day operations, because once you buy something you’ll be living with it every day for quite a while. Also ensure you have the operational groups on board during the PoC – particularly the network and endpoint folks. Implementing NAC (or something like it) impacts both these areas – often quite significantly. And the last thing you need is another group sabotaging your efforts because you didn’t line up support early in the process. Step 5: Initial Deployment/Quick Win At this point, after you have selected and bought technology (yes, we skipped a bunch of steps, including actually buying the gear), you need to roll it out. For NAC, we recommend most organizations focus on visibility initially. This provides dashboards and reports about what devices are connecting, where they are going, and what they are doing. Gradually enforcement policies for some classes of users/devices can be introduced – once you figure out where the biggest exposures are, based on real usage rather than the theoretical threat model. We favor visibility first because this is about getting a quick win. Breaking users’ ability to get onto the network and do work qualifies as a big loss. To take it a level deeper, given the sensitivity around mobile devices, a logical place to start is monitoring the mobile devices on your network. In our experience this is pretty enlightening, and will clearly drive the first set of access control policies. Alternatively you could scrutinize guest access or folks coming in on the VPN from unprotected networks. We aren’t religious about where you start, but make sure you focus on a place where you know bad stuff is happening. This way you get proof of the bad stuff and then take quick action to block it, which becomes a quick win. Then you can focus on the next area of bad stuff and build momentum for the technology and project. Wrapping up Given that most of these project have some kind of compliance driver, you also need to focus on documentation during the project. Document how you achieve some aspect of whatever compliance mandate you worry doubt. Document how you compare to the success criteria you established early on in the project. Make sure to document the support you lined up from other operational groups throughout the project. That will help when they inevitably push back on deploying the technology for some reason or other. We have spent considerable time thinking about the impact of any computing (providing access from anywhere, at any time, on any device) on how we need to protect our networks. These emerging requirements – especially in light of the avalanche of consumer-oriented mobile devices – are driving us to providing Network Access Control capabilities on our networks. Whether implementing a specific NAC device or using your existing switching and security infrastructure, you need the ability to guard against unauthorized access to your most critical information. This involves a number of choices about integrating with the existing network and security infrastructure, as well as endpoint/mobile device management, depending on the level of remediation required on out-of-policy devices. There are many potential issues regarding this integration and remediation which must be identified and addressed during the procurement process, so focus on a modest initial roll-out which both provides answers for followup and builds momentum though quick wins. It sounds easy, and on paper it is. You’ll find real

FAM is a relatively new technology, but we already see the emergence of consistent architectural models. The key components are a central management server, sensors, and connectors to the directory infrastructure. Central Management Server The core function of FAM is to monitor user activity on file repositories. While simple conceptually, this information is only sometimes available natively from the repository, and enterprises store their sensitive documents and files using a variety of different technologies. This leads to three main deployment options – each of which starts with a central management server or appliance: Single Server/Appliance: A single server or appliance serves as both the sensor/collection point and management console. This configuration is typically used for smaller deployments and when installing collection agents isn’t possible. Two-tier Architecture: This is a central management server and remote collection points/sensors. The central server may or may not monitor directly; but either way it aggregates information from remote systems, manages policies, and generates alerts. The remote collectors may use any of the collection techniques we will discuss later, and always feed data back to the central server. Hierarchical Architecture: Collection points/sensors aggregate to business-level or geographically distributed management servers, which in turn report to an enterprise management server. Hierarchical deployments are best suited for large enterprises which may have different business unit or geographic needs. They can also be configured to only pass certain kinds of data between tiers, in order to handle large volumes of information, to support privacy by unit or geography, and to support different policy requirements. Whichever deployment architecture you choose, the central server aggregates all collected data (except deliberately excluded data), performs policy-based alerting, and manages reporting and workflow. The server itself may be available in one of three flavors (or for hierarchical deployments, a combination of the three): Dedicated appliance Software/server Virtual appliance Which flavors are available depends on the vendor, but most offer at least one native option (appliance/software) and a virtual appliance. If the product supports blocking this usually handled by configuring it as a transparent bridge or in the server agent (which we will discuss about in a moment). We will discuss the central server functions in a later post. Sensors The next component is the sensors used to collect activity. Remember that this is a data-center oriented technology, so we focus on the file repositories, not the file access points (endpoints). There are three primary homes for files: Server-based file shares (Windows and UNIX/Linux) Network Attached Storage (NAS) Document Management Systems (including SharePoint) SANs are generally accessed through servers attached to a controller/logical unit or document management systems, so FAM systems focus on the file server/DMS and ignore the storage backend. FAM tools use one of three options to handle all these technologies: Network monitoring: Passive monitoring of the network outside the repository, which may be done in bridge mode or in parallel, by sniffing at a SPAN or mirror port on the local network segment. The FAM sensor or server/appliance only sniffs for relevant traffic (typically the CIFS protocol, and possibly others like WebDAV). Server agent: This is an operating system-specific agent that monitors file access on the server (usually Windows or UNIX/Linux). The agent does the monitoring directly, and does not rely on native OS audit logs. Application integration: Certain NAS products and document management systems support native auditing well beyond what’s normally provided by operating systems. In these cases, the FAM product may integrate via an agent, extension, or administrative API. The role of the sensor is to collect activity information: who accessed the file, what they did with them (open, delete, etc.), and when. The sensor should also track important information such as permission changes. Directory Integration This is technically a function of the central management server, but may involve plugins or agents to communicate with directory servers. Directory integration is one of the most important functions of a File Activity Monitor. Without it the collected activity isn’t nearly as valuable. As you’ll see when we talk about the different functions of the technology, one of the most useful is the ability to manage user entitlements and scan for things like excessive permissions. You can assume Active Directory is supported, and likely LDAP, but if you have an unusual directory server, be sure to check with the vendor before buying any FAM products. Roles and permissions change on a constant basis, so it’s important for this data flow to happen as close to real time as possible so the FAM tool knows, at all times, the actual group/role status of users. Capturing Access Controls (File Permissions) Although this isn’t a separate architecture component, all File Activity Monitors are able to capture and analyze existing file permissions (something else we will discuss later). This is done by granting administrator or file owner permissions to the FAM server or sensor, which then captures file permissions and sends them back to the management server. Changes are then synchronized in real time through monitoring, and in some cases the FAM is used to manage future privilege changes. That’s it for the base architecture; in our next post we’ll start talking about all the nifty features that run on these components. Share: