Research

On Friday the Mac Wireless hack issue exploded again after Apple PR issued a carefully worded press release. Next thing you know one of my favorite sites, The Unofficial Apple Weblog posts a headline that’s just wrong. There have been a lot of really bad posts on this topic, but John Gruber at Daring Fireball winds his way through the press and blog hype in a well reasoned article, The Curious Case of the Supposed MacBook Wi-Fi Hack. John’s reasoning is strong, but I believe we can take his assumptions in a different direction and finish with essentially the opposite results. First some full disclosure- I was at Black Hat and Defcon, talked with Maynor and Ellch, and have followed up with Maynor and SecureWorks since the event. I won’t be revealing any secret information here, but will just analyze John Gruber’s assumptions and see how his conclusions might change. John and I also emailed a bit on this issue over the weekend (he’s on vacation this week, so might not be able to respond). For those of you with short attention spans I believe that Maynor and Ellch will emerge with their reputations intact and have been trying to do the right thing from the start. If I’m wrong I’ll be the first to call myself on it and apologize, but I really don’t expect that to happen. John’s first assumption is: “What”s notable about this disclosure is that it is about the driver. We already know, just from watching the demonstration video, that it was also based on a third-party card. This means that either (a) the exploit they discovered uses neither the MacBook”s built-in card nor Mac OS X”s built-in driver; (b) the exploit they discovered works against both the third-party driver demonstrated in the video and against Apple”s standard driver, and they have inexplicably decided to post this disclaimer to explicitly describe only what is being demonstrated in the video; or (c) that the “experts” at SecureWorks do not understand the difference between a driver and a card. My money is on (a).” Let’s explore option (b), especially the last part: ‘…they have inexplicably decided to post this disclaimer to explicitly describe only what is being demonstrated in the video’ . (bold added) I propose an alternative: that they purposely posted the disclaimer to explicitly describe only what is being demonstrated in the video. Why would they do this? Not all security researchers believe in full disclosure. If you are one of these researchers and you don’t want to disclose the details of an unpatched vulnerability but want to demonstrate the class of vulnerability (device driver exploits) you might choose to demonstrate the vulnerability using an unidentified device. In the background you would notify any affected vendors and give them time to respond. If you show the attack on the built-in wireless device you instantly identify the vendor involved. An anonymous third-party card avoids this exposure. Let’s move to the next few points which focus on Brian Krebs. John states, “The reason this is notable is that if (a) is true (that the vulnerability they discovered does not apply to the standard AirPort driver software from Apple) it entirely contradicts Brian Krebs”s original and much-publicized story. Krebs wrote (emphasis added): “The video shows Ellch and Maynor targeting a specific security flaw in the Macbook”s [sic] wireless “device driver,” the software that allows the internal wireless card to communicate with the underlying OS X operating system. While those device driver flaws are particular to the MacBook – and presently not publicly disclosed – Maynor said the two have found at least two similar flaws in device drivers for wireless cards either designed for or embedded in machines running the Windows OS. Still, the presenters said they ultimately decided to run the demo against a Mac due to what Maynor called the “Mac user base aura of smugness on security.” Brian is a reporter and as such has different motivations than a security researcher. Brian posted this information and stands by it. Maynor and Ellch have followed a policy of not commenting on the potential vulnerability of native MacBook wireless drivers. Thus we have a situation where Brian reported something, but the sources won’t validate or repudiate the statement. Maynor and Ellch have yet to either confirm or deny Brian’s reporting. Why might they do this? If the vulnerability was real and they didn’t wish to disclose it until the vendor involved issued a patch. For this to be true they would have to have informed Apple (and the anonymous third-party device vendor) and said vendor wasn’t ready, for whatever reasons, to issue a patch. Since we’re only a few weeks from the initial disclosure we’re still in a reasonable timeframe. Remember, if they confirm Brian’s post they thus release enough details on the vulnerability that it could be replicated. But they haven’t denied the statement, which either indicates it’s true, Brian is wrong, or they lied. I don’t believe this is something they would lie about. Brian is now in the unenviable position of trying to justify his reporting without confirmation from his sources. While not a reporter (I’m just an analyst and blogger) I’ve come close to similar situations and they’re no fun. Next we have to look at Apple’s official response. John states: “In response to SecureWorks”s admission that their demonstration did not exploit the built-in driver, Apple on Friday released a statement regarding the supposed vulnerability. Lynn Fox, Apple”s director of Mac PR, told Macworld: “Despite SecureWorks being quoted saying the Mac is threatened by the exploit demonstrated at Black Hat, they have provided no evidence that in fact it is. To the contrary, the SecureWorks demonstration used a third party USB 802.11 device – not the 802.11 hardware in the Mac – a device which uses a different chip and different software drivers than those on the Mac. Further, SecureWorks has not shared or demonstrated any code in relation to the Black Hat-demonstrated

As previously posted I have a fair bit of experience with security screening in large facilities. With all the hype about airports these days it’s a good time to review the screening process and the role it plays in securing public areas. While one of the risks of security is believing expertise in one domain means expertise in all areas I believe large facilities/events and airports are related closely enough that we can apply the lessons of one to the other. In summary the security screening process is an effective tool at reducing risk in controlled spaces but is an extremely ineffective tool at completely eliminating risk. Screening only works when used in conjunction with other security controls both inside and outside of the area/facility being protected in a layered model. Good old defense in depth applies just as much to the physical world as the electronic one. Screening has improved a bit in U.S. airports since 9/11 but seems to be relied on too heavily and just can’t provide provide the level of protection either the public or politicians seem to demand from it. Continually increasing scrutiny during airport screenings beyond normal levels or increasing ID requirements will not significantly reduce the risk of a successful attack. At this point the best way to reduce the chances of a successful attack is to use additional security controls spread throughout the air travel system. I’ll talk a bit at how we screened at events, but if you just want to know about airports you might want to skip to the end. Back in the 90’s we used to perform physical searches on most of the people coming to concerts and sports events. Screening was just one of the many tools we used to ensure public safety at the event and it worked well for what we expected from it. Our screening met four goals: Check tickets or authenticate ID (depending on the event) Reduce unallowed items, like alcohol or cameras Reduce dangerous items like weapons/drugs Profile the person for entry (stopping extreme drunks, identifying problem children) (more after the jump…) Over the 10 years I worked event security I probably hand searched tens of thousands of people. It’s definitely not as fun as it sounds (and is surprisingly hard on the knees). We would vary screening considerably based on the risk profile of the event. A Barry Manilow show might get a quick visual check, while the latest skinhead gig would border a strip search. We didn’t use metal detectors due to a combination of cost and limited effectiveness. If all you’re looking for is metal they’re not too bad, but they suck for snagging boda bags full of Jack or vials of coke. One of the nice things about our events is that they were technically private- every ticket included a disclaimer that entry was solely at our discretion. You didn’t have to agree to a search, but we didn’t have to let you in. It seems harsh but some events can get a little rough. If we let everyone in with whatever they wanted things could get out of control in some very dangerous ways. Not for us, but for the people attending. I think it was a little worse back then, I rarely see the same kinds of brawls and injuries when I go to a show these days. Screening for any particular show was designed using a few factors: Size of the event Number of staff and experience level Nature of the event/audience Special requests from the performer or event sponsor (like no cameras) Budget A football game is huge- Folsom Stadium at the University of Colorado seats around 53,000 while the old Mile High stadium was just over 70,000. To cover an event of that size you need a large pool of cheap labor willing to work a single (or handful) of events screening at a dozen or more gates, with 5-15 screeners per gate. There’s no way to train an inconsistent group like that so we’d give them a quick briefing and place one or more experienced supervisors and regular staff with them. In this case the goal of screening is mostly to reduce the amount of booze making it’s way inside. When you convert a stadium to a concert venue that same facility might increase capacity by 30,000 while increasing the risk profile. Guns and Roses/Metallica was around 100,000. As much as we’d like otherwise we had to accept that screening at an event of that size would be only somewhat effective, so we beef up security in the event itself. We might run 50% more staff at a stadium concert than a football game. Smaller shows usually had smaller budgets. For shows of all sizes we had to try and match staff to get the attendees in as quickly as possible. Most people show up within an hour of a show starting, so we might have to hand search 3,000-10,000 people in that time and we’d balance staff and budget as best we could. For high risk shows we’d use the most experienced staff possible and rotate frequently to keep people fresh. There’s no manual for this stuff, but after working enough shows you get a good feel for the different kinds of crowds, when they’ll show up at the doors, what they’ll try to sneak in (everyone sneaks something in), and what to look for in terms of behavior and dress. You’d be surprised at which shows were the best or worst. Buffett concerts and Dead shows had plenty of fights and injuries, while speed metal shows were usually pretty tame aside from the mosh pits. But screening was only one small part of our security controls. Security typically stared in the parking lots and roads around the event, with a mix of identified staff and a few people in regular clothes. Out in the parking lots we’d look for problems, take care of the drunks, and give early verbal instructions

After posting Concerts vs. Airports: The Role and Effectiveness of Security Screening in Public Places I realized it was a tad long and I might bore some of you, so here’s the crib notes: For about ten years I worked, and eventually directed, security for large events like concerts and football games. There are some lessons we can apply to airline screening since both involve securing public spaces and large crowds: Screening is just one layer of security, but in airports it’s treated as practically the only layer. In concerts we relied more heavily on inside security to fill the gaps of screening. In concerts we used more behavioral profiling, earlier in the system. We never stopped profiling and watching once someone was inside. Technology is only good at catching certain things, and can’t catch everything. Increasing screening, but not the rest of security, will only piss people off and won’t significantly improve security. 80% or more of airport security seems to start and stop with screening. This might look good, but isn’t really secure. Computers suck at profiling and and are much more likely to catch a good guy than a bad guy. No security is perfect, and a determined and intelligent attacker could probably defeat most security where we allow public access, but by adding additional non-intrusive security controls we can rely less on screening and increase security while improving the flight experience. We need to build more layers into air transport security, not try and build a single really big wall. We have defense in depth at concerts and football games, why not airports? There’s more in the main post, but you get the idea. Share: