Research

So far in this series on tokenization guidance for protecting payment data, we have covered deficiencies in the PCI supplement, offered specific advice for merchants to reduce audit scope, and provided specific tips on what to look for during an audit. In this final post we will provide a checklist of each PCI requirement affected by tokenization, with guidance on how to modify compliance efforts in light of tokenization. I have tried to be as brief as possible while still covering the important areas of compliance reporting you need to adjust. Here is our recommended PCI requirements checklist for tokenization: PCI Requirement   Recommendation 1.2 Firewall Configuration   Token server should restrict all IP traffic – to and from – to systems specified under the ‘ground rules’, specifically: *      Payment processor *      PAN storage server (if separate) *      Systems that request tokens *      Systems that request de-tokenization This is no different than PCI requirements for the CDE, but it’s recommended that these systems communicate only with each other. If the token server is on site, Internet and DMZ access should be limited to communications with payment processor.   2.1 Defaults Implementation for most of requirement 2 will be identical, but section 2.1 is most critical in the sense there should be no ‘default’ accounts or passwords for the tokenization server. This is especially true for systems that are remotely managed or have remote customer-care options. All PAN security hinges on effective identity and access management, so establishing unique accounts with strong pass-phrases is essential.   2.2.1 Single function servers 2.2.1 bears mention both from a security standpoint, as well as protection from vendor lock-in. For security, consider an on-premise token server as a standalone function, separate and distinct from applications that make tokenization and de-tokenization requests.   To reduce vendor lock-in, make sure the token service API calls or vendor supplied libraries used by your credit card processing applications are sufficiently abstracted to facilitate switching token services without significant modifications to your PAN processing applications.   2.3 Encrypted communication You’ll want to encrypt non-console administration, per the specification, but all API calls to the token service as well. It’s also important to consider, when using multiple token servers to support failover, scalability or multiple locations that all synchronization occurs over encrypted communications, preferably a dedicated VPN with bi-directional authentication.   3.1 Minimize PAN storage The beauty of tokenization is it’s the most effective solution available for minimizing PAN storage. By removing credit card numbers from every system other than the central token server, you cut the scope of your PCI audit. Look to tokenize or remove every piece of cardholder data you can, and keeping it in the token server as well. You’ll still meet business, legal and regulatory requirements while improving security.   3.2 Authentication data Tokenization does not circumvent this requirement; you must remove sensitive authentication data per sub-sections 3.2.X 3.3 Masks Technically you are allowed to preserve the first six (6) digits and the last four (4) digits of the PAN. However, it’s our recommendation that you examine your business processing requirements to determine if you can fully tokenize the PAN, or at a minimum, only preserve the last four digits for customer verification. The number of possible tokens you can generate with the remaining six digits is too small for many merchants to generate quality random numbers. Please refer to ongoing public discussions on this topic for more information.   When using a token service from your payment processor that you ask for single-use tokens to avoid possible cases of cross-vendor fraud.   3.4 Render PAN unreadable One of the principle benefits of tokenization is that it renders PAN unreadable. However, auditing environments with tokenization requires two specific changes: 1.     Verify that PAN data is actually swapped for tokens in all systems. 2.     For on-premise token servers, verify that the token server adequately encrypts stored PAN, or offers an equivalent form of protection, such as not storing PAN data*.   We do not recommend hashing as it offers poor PAN data protection. Many vendors’ store hashed PAN values in the token database as a means of speedy token lookup, but while common, it’s a poor choice. Our recommendation is to encrypt PAN data, and as many use databases to store information, we believe table, column or row level within the token database is your best bet. Use of full database or file layer encryption can be highly secure, but most solutions offer no failsafe protection when database or token admin credentials are compromised.   We acknowledge our recommendations differ from most, but experience taught us to err on the side of caution when it comes to PAN storage.   *Select vendors offer one-time pad and codebooks options that don’t require PAN storage.   3.5 Key management Token server encrypt the PAN data stored internally, so you’ll need to verify the supporting key management system as best you can. Some token servers offer embedded key management, while others are designed to leverage your existing key management services.   There are very few people who can adequately evaluate key management systems to ensure they are really secure, but at the very least, you can check that the vendor is using industry standard components, or has validated their implementation with a 3rd party. Just make sure they are not storing the keys in the token database unencrypted. It happens.   4.1 Strong crypto on Internet communications As with requirement 2.3, when using multiple token servers to support failover, scalability and/or multi-region support that all synchronization occurs over encrypted communications, preferably a dedicated VPN with bi-directional authentication.   6.3 Secure development One site or 3rd party token servers will both introduce new libraries and API calls into your environment. It’s critical that your development process includes the validation that what you put into production is secure. You can’t take the vendor’s word for it – you’ll need to validate that all defaults, debugging code and API calls are secured. Ultimately

As we resume our tour of advanced use cases for Network Security Analysis, it’s time to consider malware analysis. Of course most successful attacks involve some kind of malware at some point during the attack. If only just to maintain a presence on the compromised device, some kind of bad stuff is injected. And once the bad stuff is on a device, it’s very very hard to get rid of it – and even harder to be sure. Most folks (including us) recommend you just re-image the device, as opposed to trying to clean the malware. This makes it even more important to detect malware as quickly as possible and (hopefully) block it before a user does something stupid to compromise their device. There are many ways to detect malware, depending on the attack vector, but a lot of what we see today is snuck through port 80 as web traffic. Sure, dimwit users occasionally open a PDF or ZIP file from someone they don’t know, but more often it’s a drive-by download, which means it comes in with all the other web traffic. So we have an opportunity to detect this malware when it enters the network. Let’s examine two situations, one with a purpose-built device to protect against web malware, and another where we’re analyzing malware directly on the network analysis platform. Detecting Malware at the Perimeter As we’ve been saying throughout this series, extending data collection and capture beyond logs is essential to detecting modern attacks. One advantage of capturing the full network packet stream at the ingress point of your network is that you can check for known malware and alert as they enter the network. This approach is better than nothing but it has two main issues: Malware sample accuracy: This approach requires accurate and comprehensive malware samples already loaded into the device to detect the attack. We all know that approach doesn’t work well with endpoint anti-virus and is completely useless against zero-day attacks, and this approach has the same trouble. No blocking: Additionally, once you detect something on the analysis platform, your options to remediate are pretty limited. Alerting on malware entering the network is useful, but blocking it is much better. Alternatively, a new class of network security device has emerged to deal with this kind of sneaky malware, by exploding the malware as it enters the network to understand the behavior of inbound sessions. Again, given the prevalence of unknown zero-day attacks, the ability to classify known bad behavior and see how a packet stream actually behaves can be very helpful. Of course no device is foolproof, but these devices can provide earlier warning of impending problems than traditional perimeter network security controls. Using these devices you can also block the offending traffic at the perimeter if it is detected in time, reducing the likelihood of device compromise. But you can’t guarantee you will catch all malware, so you must figure out the extent of t he compromise. There is also a more reactive approach: analyzing outbound traffic to pinpoint known command and control behavior and targets, which usually indicating a compromised device. At this point the device is already pwned, so you need to contain the damage. Either way, you must figure out exactly what happened and whether you need to sound the alarm. Containing the Damage Based on the analysis on the perimeter, we know both the target device and the originating network address. With our trusty analysis platform we can then figure out the extent of the damage. Let’s walk through the steps: Evaluate the device: First you need to figure out if the device is compromised. Your endpoint protection suite might not be able to catch an advanced attack, so search your analysis platform (and logging system) to find any configuration changes made on the device, and look for any strange behavior – typically through network flow analysis. If the device is still clean all the better. But we will assume it’s not. Profile the malware: Now you know the device is compromised, you need to figure out how. Sure you could just wipe it, but that eliminates the best opportunity to profile the attack. The network traffic and device information enable your analysts to piece together exactly what the malware does, replay the attack to confirm, and profile its behavior. This helps figure out how many other devices have been compromised, because you know what to look for. Determine the extent of the damage: The next step is to track malware proliferation. You can search the analysis platform to look for the malware profile you built in the last step. This might mean looking for communication with the external addresses you identified, identifying command and control patterns, or watching for the indicative configuration changes; but however you proceed, having all that data in one place facilitates identifying compromised devices. Watch for the same attack: You know the saying: “Fool me once, shame on you. Fool me twice, shame on me.” Shame on you if you let the same attack succeed on your network. Add rules to detect and block attacks you have seen for the future. We have acknowledged repeatedly that security professionals get no credit for blocking attacks, but you certainly look like a fool if you get compromised repeatedly by the same attack. You are only as good as your handling of the latest attack. So learn from these attacks; the additional data collection capabilities of network security analysis platforms can give you an advantage, both for containing the damage and for ensuring it doesn’t happen again. As we wrap up this Applied Network Security Analysis series early next week, we will examine the use case of confirming a breach actually happened, and then revisit the key points to solidify our case for capturing network traffic as a key facet of your detection capabilities. Share:

I wouldn’t say I’m a control freak, but I am definitely “control aligned”. If something is important to me I like to know what’s going on under the hood. I also hate to depend on someone else for something I’m capable of. So I have no problem trusting my accountant to keep me out of tax jail, or hiring a painter for the house, but there is a long list of things I tend to overanalyze and have trouble letting go of. Pretty damn high up that list is the Securosis Nexus. I have been programming as a hobby since third grade, and for a while there in the early days of web applications it was my full time profession. I don’t know C worth a darn, but I was pretty spiffy with database design and my (now antiquated) toolset for building web apps. I still code when I can, but it’s more like home repair than being a general contractor. When Mike, Adrian, and I came up with the idea for the Nexus I did all the design work. From the UI sketches we sent to the visual designers to the features and logic flow. Not that I did it all alone, but I took point, and I’m the one who interfaces with our contractors. Which is where I’m learning how to let go. The hard way. I have managed (small) programming teams before but this is my first time on the hiring side of the contractor relationship. It’s also the first time I haven’t written any significant amount of code for something I’m pretty much betting my future on (and the future of my partners and our families). Our current contractor team is great. Among other things they suggested an entirely new architecture for the backend that is far better than my initial plans and our PoC code. I wish they would QA a little better (hi guys!), and we don’t always see things the same way, but I’m damn happy with the product. But it’s extremely hard for me to rely on them. For example, today I wanted to change how a certain part of the system functioned (how we handle internal links). I know what needs to be done, and even know generally what needs to happen within the code, but I realized I would probably just screw it up. And it would take me a few hours (to screw up), while they can sort it all out in a fraction of the time. I don’t know why this bothers me. Maybe it’s knowing that I’ll see a line item on an invoice down the road. But it’s probably some deep-seated need to feel I’m in control and not dependant on someone else for something so important. But I am. And I need to get used to it. On to the Summary: Webcasts, Podcasts, Outside Writing, and Conferences Me (Rich) in a DLP video I for Trend Micro. I really liked the video crew on this one and the quality shows. I may need to get myself a Canon DSLR for our future Securosis videos instead of our current HD camcorder. I also wrote up how to recover lost iCloud data based on my own serious FAIL this week. Favorite Securosis Posts Mike Rothman: Virtual USB? Not.. Adrian has it right here. Even though it’s more secure to carry (yet another) device, users won’t do it. They want everything on their smartphone, and they will get it. It’s just a matter of when, and at what cost (in terms of security or data loss). Adrian Lane: How Regular Folks See Online Safety. Lately news items are right out of Theater of the Absurd: Security Tragicomedy. Rich: Tokenization Guidance: Audit Advice. Adrian is really building the most definitive guide out there. Other Securosis Posts Incite 11/2/2011: Be Yourself. Conspiracy Theories, Tin Foil Hats, and Security Research. Applied Network Security Analysis: The Advanced Security Use Case. Applied Network Security Analysis: The Forensics Use Case. Favorite Outside Posts Mike Rothman: 3 Free Tools to Fake DNS Responses for Malware Analysis. This is a good tip for testing, but also critical for understanding the tactics adversaries will use against you. Adrian Lane: The Chicago Way. Our own Dave Lewis does the best job in the blogsphere at explaining what the heck is going on with the Anonymous / Los Zetas gang confrontation. James Arlen: Harvard Stupid. Two posts in one – interesting financial story tailed by an excellent example of how security should be implemented from a big picture view. If you run IT security for your company, read this! Rich: Kevin Beaver on why users violate policies. I don’t agree with the lazy comment though – it’s not being lazy if your goal is to get your job done and you deal with something in the way. Research Reports and Presentations Fact-Based Network Security: Metrics and the Pursuit of Prioritization. Tokenization vs. Encryption: Options for Compliance. Security Benchmarking: Going Beyond Metrics. Understanding and Selecting a File Activity Monitoring Solution. Database Activity Monitoring: Software vs. Appliance. React Faster and Better: New Approaches for Advanced Incident Response. Measuring and Optimizing Database Security Operations (DBQuant). Network Security in the Age of Any Computing. Top News and Posts UK Cops Using Fake Mobile Phone Tower to Intercept Calls, Shut Off Phones. Malaysian CA Digicert Revokes Certs With Weak Keys, Mozilla Moves to Revoke Trust. Four CAs Have Been Compromised Since June. Hackers attacked U.S. government satellites. How Visa Protects Your Data. Exposing the Market for Stolen Credit Cards Data. ‘Nitro’ Cyberespionage Attack Targets Chemical, Defense Firms. Blog Comment of the Week This week we are redirecting our donation to support Brad “theNurse” Smith. This week’s best comment goes to Zac, in response to Conspiracy Theories, Tin Foil Hats, and Security Research. I personally think that the problem with the media hype is that it seems to distract more than inform. The overall result being that you end up with “experts” arguing over inconsequential

Last week I was invited to speak at Kennesaw State University’s annual cybersecurity awareness day. They didn’t really give me much direction on the topic, so I decided to give my Happyness presentation. I figured there would be students and other employees who could benefit from my journey from total grump to fairly infrequent grump, and a lot of the stuff I’ve learned along the way. One of my key lessons is to accept the way I am and stop trying to be someone else. Despite my public persona I like (some) people. Just not many, and in limited doses. I value and need my solitary time and I have designed a lifestyle to embrace that. I say what I think, and know I can be blunt. Don’t ask me a question if you don’t want the answer. Sure, I have mellowed over the years, but ultimately I am who I am, and my core personality traits are unlikely to change. The other thing I have realized is the importance of managing expectations. For example, I was at SecTor CA a few weeks back, and at the beginning of my presentation on Cloud Security (WMV file), I mentioned the Internet with a snarky, “You know, the place were pr0n is.” (h/t to Rich – it’s his deck). There was a woman sitting literally in the front row who blurted out, “That’s totally inappropriate.” I immediately stopped my pitch, because this was a curious comment. I asked the woman what she meant. She responded that she didn’t think it was appropriate to mention pr0n on a slide in a conference presentation. Yeah, I guess she doesn’t get to many conferences. But it wasn’t something I was going to gloss over. So I responded: “Oh you think so, then this may not be the session for you.” Yes, I really said that, much to the enjoyment of everyone else in the room. I figured given the rest of the content and my presentation style that this wasn’t going to end well. There was no reason for her to spend an hour and be disappointed. To her credit, she got up and found another session, which was the best outcome for both of us. Earlier in my career, I would have let it go. I would probably have adapted my style a bit to be less, uh, offensive. I would have gotten the session done, but it wouldn’t have been my best effort. Now I just don’t worry about it. If you don’t like my style, leave. If you don’t think I know what I’m talking about, leave. If you don’t like my blog posts, don’t read them. It’s all good. I’m not going to feel bad about who I am. Which philosophy is directly from Steve Jobs. “Your time is limited, so don’t waste it living someone else’s life.” I have got lots of problems, but trying to be someone else isn’t one of them. For that I’m grateful. So just be yourself, not who they want you to be. That’s the only path to make those fleeting moments of happiness less fleeting. -Mike Photo credits: “Just be Yourself” originally uploaded by Akami Incite 4 U Keeping tabs on theNurse: I know Brad “theNurse” Smith isn’t familiar to most of you, but if you have been to a major security conference, odds are you have seen him and perhaps met him. I first met Brad 5+ years ago when we worked as Black Hat room proctors together, and have since seen him all over the place. Last week Brad suffered a serious stroke while delivering a presentation at the Hacker Halted conference in Miami, and he still hasn’t regained consciousness. You can get updates on Brad over at the social-engineer.org site, and can leave donations if you want. Maybe I’m identifying a bit too much after my recent health scare on the road, but we feel terrible for Brad and his wife and all of us at Securosis wish them the best. We are also putting our money where our mouths are, and directing (and increasing) our Friday Summary donation his way this week. – RM The weakest link? Your people… I just love stories of social engineering. Yes, there are some very elegant technical attacks, but they seem so much harder than just asking for access to the stuff you need. Like a wiring closet or conference room. Why pick the lock on the door when they’ll just open when you knock? Kai Axford had a great video (WMV) of actually putting his own box into a pen test client’s wiring closet – with help from the network admin – in his SecTor CA presentation. And NetworkWorld has a good story on social engineering, including elegant use of a tape measure. But it’s not like we haven’t seen this stuff before. On my golf trip, we stumbled across Beverly Hills Cop on a movie channel and Axel Foley is one of the best social engineers out there. – MR Token gesture: 403 Labs QSA and PCI columnist Walt Conway noted a major change to the PCI Special Interest Groups (SIGs) this year. The “participating organizations” – a group comprised mostly of the merchants who are part of the PCI Council – will get the deciding vote on which SIGs get to provide the PCI Council advice. Yes, they get a vote on what topics get the option of community guidance. The SIGs do a lot of the discovery and planning work that goes into the guidance ultimately published by the PCI Council – end to end encryption is one example. Unless, of course, someone like Visa objects to the SIG’s guidance, in which case the PCI Council squashes it like a bug – as they did with tokenization. This olive branch is nice, but it’s a token minuscule gesture. – AL Job #1: Keep head attached to body: I joke a lot during presentations about the importance of a public execution

I remember very clearly the day I vowed to stop watching local news. I was sitting at home cooking dinner or something, when a teaser report of a toddler who died after being left in a car in the heat aired during that “what we’re covering tonight” opening to the show. It wasn’t enough to report the tragedy – the reporter (a designation she surely didn’t deserve) seemed compelled to illustrate the story by locking a big thermometer in the car, to be pulled out during the actual segment. Frankly, I wanted to vomit. I have responded to more than a few calls involving injured or dead children, and I was disgusted by the sensationalism and desperate bid for ratings. With rare exceptions, I haven’t watched local news since then; I can barely handle cable news (CNN being the worst – I like to say Fox is right, MSNBC left, and CNN stupid). But this is how a large percentage of the population learns what’s going on outside their homes and work, so ‘news’ shows frame their views. Local news may be crap, but it’s also a reflection of the fears of society. Strangers stealing children, drug assassins lurking around every corner, and the occasional cancer-causing glass of water. So I wasn’t surprised to get this email from a family member (who found it amusing): Maybe you have seen this, but thought I would send it on anyway. SCARY.. This is a MUST SEE/ READ. If you have children or grandchildren you NEED to watch this. I had no idea this could happen from taking pictures on the blackberry or cell phone. It’s scary. http://www.youtube.com/watch?v=N2vARzvWxwY Crack open a cold beer and enjoy the show… it’s an amusing report on how frightening geotagged photos posted online are. I am not dismissing the issue. If you are, for example, being stalked or dealing with an abusive spouse, spewing your location all over the Internet might not be so smart. But come on people, it just ain’t hard to figure out where someone lives. And if you’re a stalking victim, you need better sources for guidance on protecting yourself than stumbling on a TV special report or the latest chain mail. But there are two reasons I decided to write this up (aside from the lulz). First, it’s an excellent example of framing. Despite the fact that there is probably not a single case of a stranger kidnapping due to geotagging, that was the focus of this report. Protecting your children is a deep-seated instinct, which is why so much marketing (including local news, which is nothing but marketing by dumb people) leverages it. Crime against children has never been less common, but plenty of parents won’t let their kids walk to school “because the world is different” than when they grew up. Guess what: we are all subject to the exact same phenomenon in IT security. Email is probably one of the least important data loss channels, but it’s the first place people install DLP. Not a single case of fraud has ever been correlated with a lost or stolen backup tape, but many organizations spend multiples more on those tapes than on protecting web applications. Second, when we are dealing with non-security people, we need to remember that they always prioritize security based on their own needs and frame of reference. Policies and boring education about them never make someone care about what you care about as a security pro. This is why most awareness training fails. To us this report is a joke. To the chain of people who passed it on, it’s the kind of thing that freaks them out. They aren’t stupid (unless they watch Nancy Grace) – they just have a different frame of reference. Share:

In this portion of our Tokenization Guidance series I want to offer some advice to auditors. I am addressing both internal auditors going through one of the self assessment questionnaires, as well as external auditors validating adherence to PCI requirements. For the most part auditors follow PCI DSS for the systems that process credit card information, just as they always have. But I will discuss how tokenization alters the environment, and how to adjust the investigation process in the select areas where tokenization systems supplants PAN processing. At the end of this paper, I will go section by section through the PCI DSS specification and talk about specifics, but here I just want to provide an overview. So what does the auditor need to know? How does it change discovery processes? We have already set the ground rules: anywhere PAN data is stored, applications that make tokenization or de-tokenization requests, and all on-premise token servers require thorough analysis. For those systems, here is what to focus on: Interfaces & APIs: At the integration points (APIs and web interfaces) for tokenization and de-tokenization, you need to review security and patch management – regardless of whether the server is in-house or hosted by a third party. The token server vendor should provide the details of which libraries are installed, and how the systems integrate with authentication services. But not every vendor is great with documentation, so ask for this data if they failed to provide it. And merchants need to document all applications that communicate with the token server. This encompasses all communication, including token-for-PAN transactions, de-tokenization requests, and administrative functions. Tokens: You need to know what kind of tokens are in use – each type carries different risks. Token Storage Locations: You need to be aware of where tokens are stored, and merchants need to designate at least one storage location as the ‘master’ record repository to validate token authenticity. In an on-premise solution this is the token server; but for third-party solutions, the vendor needs to keep accurate records within their environment for dispute resolution. This system needs to comply fully with PCI DSS to ensure tokens are not tampered with or swapped. PAN Migration: When a tokenization service or server is deployed for the first time, the existing PAN data must be removed from where it is stored, and replaced with tokens. This can be a difficult process for the merchant and may not be 100% successful! You need to know what the PAN-to-token migration process was like, and review the audit logs to see if there were issues during the replacement process. If you have the capability to distinguish between tokens and real PAN data, audit some of the tokens as a sanity check. If the merchant hired a third party firm – or the vendor – then the service provider supplies the migration report. Authentication: This is key: any attacker will likely target the authentication service, the critical gateway for de-tokenization requests. As with the ‘Interfaces’ point above: pay careful attention to separation of duties, least privilege principle, and limiting the number of applications that can request de-tokenization. Audit Data: Make sure that the token server, as well as any API or application that performs tokenization/de-tokenization, complies with PCI section Requirement 10. This is covered under PCI DSS, but these log files become a central part of your daily review, so this is worth repeating here. Deployment & Architecture: If the token server is in-house or managed on-site you will need to review the deployment and system architecture. You need to understand what happens in the environment if the token server goes down, and how token data is synchronized being multi-site installations. Weaknesses in the communications, synchronization, and recovery processes are all areas of concern; so the merchant and/or vendors must document these facilities and the auditor needs to review. Token Server Key Management: If the token server is in-house or managed on site, you will need to review key management facilities, because every token server encrypts PAN data. Some solutions offer embedded key management while others use external services, but you need to ensure this meets PCI DSS requirements. For non-tokenization usage, and systems that store tokens but do not communicate with the token server, auditors need to conduct basic checks to ensure the business logic does not allow tokens to be used as currency. Tokens should not be used to initiate financial transactions! Make certain that tokens are merely placeholders or surrogates, and don’t work act as credit card numbers internally. Review select business processes to verify that tokens don’t initiate a business process or act as currency themselves. Repayment scenarios, chargebacks, and other monetary adjustments are good places to check. The token should be a transactional reference – not currency or a credit proxy. These uses lead to fraud; and in the event of a compromised system, might be used to initiate fraudulent payments without credit card numbers. The depth of these checks varies – merchants filling out self-assessment questionnaires tend to be more liberal in interpreting of the standard than top-tier merchants and the have external auditors combing through their systems. But these audit points are the focus for either group. In the next post, I will provide tables which go point by point through the PCI requirements, noting how tokenization alters PCI DSS checks and scope. Share:

It seems far too much of security research has become like Mel Gibson in “Conspiracy Theory.” Unbalanced, mostly crazy, but not necessarily wrong. But we created this situation, so we have to deal with it. I’m reacting to the media cycle around the Duqu virus, or Son of Stuxnet, identified by F-Secure (among others). You see, no one is interested in product news anymore. No one cares about the incremental features of a vendor widget. They don’t care about success stories. The masses want to hear about attacks. Juicy attacks that take down nuclear reactors. Or steal zillions of dollars. Or result in nudie pictures of celebrities stolen from their computers or cell phones. That’s news today, and that’s why vendor research teams focus on giving the media news, rather than useful information. It started with F-Secure claiming that Duqu was written by someone with access to the Stuxnet source code. Duqu performs reconnaissance rather than screwing with centrifuges, but their message was that this is a highly sophisticated attack, created by folks with Stuxnet-like capabilities. The tech media went bonkers. F-Secure got lots of press, and the rest of the security vendors jumped on – trying to credit, discredit, expand, or contract F-Secure’s findings – anything that would get some press attention. Everyone wanted their moment in the sun, and Duqu brought light to the darkness. But here’s the thing. Everyone saying Duqu and Stuxnet were related in some way might have been wrong. The folks at SecureWorks released research a week later, making contrary claims and disputing any relation beyond some coarse similarities in how the attacks inject code (using a kernel driver) and obscure themselves (encryption and signing using compromised certificates). The media went bonkers again. Nothing like a spat between researchers to drive web traffic to the media. So who is right? That is actually the wrong question. It really doesn’t matter who is right. Maybe Duqu was done by the Stuxnet guys. Maybe it wasn’t. Ultimately, though, to everyone aside from page-whoring beat reporters who benefit from another media cycle, who’s right and who’s wrong about Duqu’s parentage aren’t relevant. The only thing that matters is that you, as a security professional, understand the attack; and have controls in place to protect against it. Or perhaps not – analyzing the attack and accepting its risk is another legitimate choice. This is how the process is supposed to work. A new threat comes to light, and the folks involved early in the cycle draw conclusions about the threat. Over time other researchers do more work and either refute or confirm the original claims. The only thing different now is that much of this happens in public, with the media showing how the sausage is made. And it’s not always pretty. But success in security is about prioritizing effectively, which means shutting out the daily noise of media cycles and security research. Not that most security professionals do anything but fight fires all day anyway. Which means they probably don’t read our drivel either… Photo credit: “Tin Foil Hat” originally uploaded by James Provost Share:

The forensics use case we discussed previously is about taking a look at something that already happened. You presume the data is already lost, the horse is out of the barn, and Pandora’s Box is open. But what if we tried to look at some of these additional data types in terms of making security alerts better, with the clear goal of reducing the window between exploit and detection: reacting faster? Can we leverage something like network full packet capture to learn sooner when something is amiss and to improve security? Yes, but this presents many of the same challenges as using log-based analysis to detect what is going on. You still need to know what you are looking for, and an analysis engine that can not only correlate behavior across multiple types of logs, but also analyze a massive amount of network traffic for signs of attack. So when we made the point in Collection and Analysis that these Network Security Analysis platforms need to be better SIEMs than a SIEM, this is what we were talking about. Pattern Matching and Correlation Assuming that you are collecting some of these additional data sources, the next step is to turn said data into actionable information, which means some kind of alerting and correlation. We need to be careful when using the ‘C’ word (correlation), given the nightmare most organizations have when they try to correlate data on SIEM platforms. Unfortunately the job doesn’t get any easier when extending the data types to include network traffic, network flow records, etc. So we continue to advocate a realistic and incremental approach to analysis. Much of this approach was presented (in gory detail) in our Network Security Operations Quant project. Identify high-value data: This is key – you probably cannot collect from every network, nor should you. So figure out the highest profile targets and starting with them. Build a realistic threat model: Next put on your hacker hat and build a threat model for how you’d attack that high value data. It won’t be comprehensive but that’s okay. You need to start somewhere. Figure out how you would attack the data if you needed to. Enumerate those threats in the tool: With the threat models, design rules to trigger based on the specific attacks you are looking for. Refine the rules and thresholds: The only thing we can know for certain is that your rules will be wrong. So you will go through a tuning process to hone in on the types of attacks you are looking for. Wash, rinse, repeat: Add another target or threat and build more rules as above. With the additional traffic analysis you can look for specific attacks. Whether it’s looking for known malware (which we will talk about in the next post), traffic destined for a known command and control network, or tracking a buffer overflow targeted at an application residing in the DMZ, you get a lot more precision in refining rules to identify what you are looking for. Done correctly this reduces false positives and helps to zero in on specific attacks. Of course the magic words are “done correctly”. It is essential to build the rule base incrementally – test the rules and keep refining the alerting thresholds – especially given the more granular attacks you can look for. Baselining The other key aspect of leveraging this broader data collection capability is understanding how baselines change from what you may be used to with SIEM. Using logs (or more likely NetFlow), you can get a feel for what is normal behavior and use that to kickstart your rule building. Basically, you assume what is happening when you first implement the system is what should be happening, and alert if something varies too far from that normal. That’s not actually a safe assumption but you need to start somewhere. As with correlation this process is incremental. Your baselines will be wrong when you start, and you adjust them over time based with operational experience responding to alerts. But the most important step is the start, and baselines help to get things going. Revisiting the Scenario Getting back to the scenario presented in the Forensics use case, how would some of this more pseudo-real-time analysis help reduce the window between attack and detection? To recap that scenario briefly, a friend at the FBI informed you that some of your customer data showed up as part of a cybercrime investigation. Of course by the time you get that call it is too late. The forensic analysis revealed an injection attack enabled by faulty field validation on a public-facing web app. If you were looking at network full packet capture, you might find that attack by creating a rule to look for executables entered into the form fields of POST transactions, or some other characteristic signature of the attack. Since you are capturing the traffic on the key database segment, you could establish a content rule looking for content strings you know are important (as a poor man’s DLP), and alert when you see that type of data being sent anywhere but the application servers that should have access to it. You could also, for instance, set up alerts on seeing an encrypted RAR file on an egress network path. There are multiple places you could detect the attack if you know what to look for. Of course that example is contrived and depends on your ability to predict the future, figuring out the vectors before the attack hits. But at lot of this discipline is based on a basic concept: “Fool me once, shame on you. Fool me twice, shame on me.” Once you have seen this kind of attack – especially if it succeeds – make sure it doesn’t work again. It’s a bit of solving yesterday’s problems tomorrow, but many security attacks use very similar tactics. So if you can enumerate a specific attack vector based on what you saw, there is an excellent

Secure USB devices – ain’t they great? They offer us the ability to bring trusted devices into insecure networks, and perform trusted operations on untrusted computers. If I could drink out of one, maybe it would be the holy grail. Services like cryptographic key management, identity certificates and mutual authentication, sensitive document storage, and a pr0nsafe web browser platform. But over the last year, as I look at the mobile computing space – the place where people will want to use secure USB features – the more I think the secure USB market is in trouble. How many of you connect a USB stick to your Droid phone? How about your iPad? My point is that when you carry your smart device with you, you are unlikely to carry a secure USB device with you as well. The security services mentioned above are necessary, but there has been little integration of these functions into the devices we carry. USB hardware does offer some security advantages, but USB sticks are largely part of the laptop model (era) of mobile computing, which is being marginalized by smart phones. Secure on-line banking, go-anywhere data security, and “The Key to the Cloud” are clever marketing slogans. Each attempts to reposition the technology to gain user preference – and fails. USB sticks are going the way of the zip drive and the CD – the need remains but they are rapidly being marginalized by more convenient media. That’s really the key: the security functions are strategic but the medium is tactical. So where does the Secure USB market segment go? It should go with the users are: embrace the new platforms. And smart device users should look for these security features embedded in their mobile platforms. Just because the media is fading does not mean the security features aren’t just as important as we move on to the next big thing. These things all tend to cycles, but the current strong fashion is to get “an app for that” rather than carry another device. Lack of strong authenication won’t make users carry and use laptops rather than phones. It is unclear why USB vendors have been so slow to react, but they need to untie themselves from their fading medium to support user demand. I am not saying secure USB is dead, but saying the vendors need to provide their core value on today’s relevant platforms. Share:

Most organizations don’t really learn about the limitations of event logs, until forensic investigators hold up their hands and explain they know what happened, but aren’t really sure how. Huh? How could that happen? It’s pretty simple: logs are a backward-looking indicator. They can help you piece together what happened, but you can only infer how. In a forensic investigation inferring anything is suboptimal. You want to know, especially given the needs to isolate the root cause of the attack and to establish remediations to ensure it doesn’t happen again. So we need to look at additional data sources to fill in gaps in what the logs tell you. Let’s take a look at a simplified scenario to illuminate the issues. We’ll look at the scenario both from the standpoint of a log-only analysis and then with a few other data sources added. For a more detailed incident response scenario, check out our React Faster and Better paper. The Forensic Limitations of Logs It’s the call you never want to get. The Special Agent on the other end of the line called to give you a heads-up: they found some of your customer data as part of another investigation into some cyber-crime activity that helps fund a domestic terrorist ring. Normally the Feds aren’t interested in giving you a heads-up until their investigation is done, but you have a relationship with this agent from your work together in the local InfraGard chapter. So he did you a huge favor. The first thing you need to do is figure out what was lost and how. To the logs! You aren’t sure how it happened, but you see some strange log records indicating changes on a application server in the DMZ. Given the nature of the data your agent friend passed along, you check the logs on the database server where that data resides as well. Interestingly enough, you find a gap in the logs on the database server, where your system collected no log records for a five-minute period a few days ago. You aren’t sure exactly what happened, but you know with reasonable certainty that something happened. And it probably wasn’t good. Now you work backwards and isolate the additional systems compromised as the attackers made their way through the infrastructure to reach their target. It’s pretty resource intensive, but by searching in the log manager you can isolate devices with gaps in their logs during the window you identified. The attackers were pretty effective, taking advantage of unpatched vulnerabilities (Damn, Ops!) and covering their tracks by turning off logging where necessary. At this point you know the attack path, and at least part of what was stolen, thanks to the FBI. Beyond that you are blind. So what can you do to make sure you aren’t similarly suprised somewhere down the line? You can set the logging system to alert if you don’t get any log records from critical assets in any 2-minute period. Again, this isn’t perfect and will result in a bunch more alerts, but at least you’ll know something is amiss before the FBI calls. With only log data you can identify what was attacked but probably not how the attack happened. Forensics Driven by Broader Data Let’s take a look at an alternative scenario with a few other data sources such as full network packet capture, network flow records, and configuration files. Of course it is still a bad day when you get the call from your pal the Special Agent. Of course Applied Network Security Analysis cannot magically make you omniscient, but how you investigate breaches changes. You still start with the logs on the perimeter server and identify the device that served as the attacker’s initial foothold. But you’ve implemented the Full Packet Capture Sandwich architecture described in the last post, so you are capturing the network traffic in your DMZ. You proceed to the network analysis console (using the full packet capture stream) and search all the traffic to and from the compromised server. Most sessions to that server are typical – standard application traffic. But you find some reconnaissance, and then something pretty strange: an executable injected into the server via faulty field validation on the web app (Damn, Developers!). Okay, this confirms the first point of exploit. Next we go to the target (keeping in mind what data was compromised) and do a similar analysis. Again, with our full packet capture sandwich in place, we captured traffic to/from the database server as well. As in the log-only scenario, we pinpoint the time period when logging was turned off, then perform a search in our analysis console to figure out what happened during that 5-minute period on that segment. Yep, a privileged account turned off logging on the database server and added an admin account to the database. Awesome. Using that account, the attacker dumped the database table and moved the data to a staging server elsewhere on your network. Now you know which data was taken, but how? You aren’t capturing all the traffic on your network (infeasible), so you have some blind spots, but with your additional data sources you are able to pinpoint the attack path. The NetFlow records coming from the compromised database server show the path to the staging server. The configuration records from the staging server indicate what executables were installed, which enabled the attacker to package and encrypt the payload for exfiltration. Further analysis of the NetFlow data shows the exfiltration, presumably to yet another staging server on another compromised network elsewhere. It’s not perfect, because you are figuring out what already happened. But now you can get back to your FBI buddy with a lot more information about what tactics the attacker used, and maybe even evidence that might be helpful in prosecution. Can’t Everyone Get Along? Clearly this is a simplified scenario that perfectly demonstrates the need to collect additional data sources to isolate the root cause and attack path of any