ForewordThe following is Bluetooth Hunter's Guide, as it appeared on page 6 of 2600 issue 29:2. This is a pretty straightforward guide that introduces the user to a number of Bluetooth scanning tools and methods, including my pet project Bluelog.
The original ASCII text file for this document can be found in the "Text" section of the Downloads page.
IntroductionSince the publication of my article “Bluetooth Hacking Primer” in 27:1, I have received quite a few questions from readers who were curious about the practicality of Bluetooth attacks in general. One of the most common questions is, understandably, how many vulnerable Bluetooth devices are out there in the first place?
That's not quite as straight forward a question as it might seem. It’s important to realize that not all discoverable devices are vulnerable to attack, nor are all vulnerable devices discoverable (more on that later). Discoverable devices are simply that, devices which we can easily detect. If nothing else, scanning for discoverable devices is useful for determining the density of Bluetooth devices in a given area. If an attacker can establish where the most users of Bluetooth devices are, they will know where to focus their efforts.
A lot of modern devices aren't discoverable at all unless specifically enabled, and even then, they usually only stay discoverable for a few minutes. While it's great that newer hardware and mobile operating systems are taking a more pragmatic approach to Bluetooth security, it doesn't do anything for the millions and millions of older devices already out in the wild. Of course, not all manufacturers are so enlightened either, so there are still some new devices that are shipping with questionable default policies. The end result is that there are still many Bluetooth devices happily announcing their presence to anyone who cares to listen.
This particular article will focus on the search for, and identification of, Bluetooth devices on a large scale. We’ll start by looking at the best hardware for long-range omnidirectional detection, and then talk about some different software options and techniques available. I’ll also be discussing some of the data I have personally collected to give you an idea of what you should expect in terms of number of results and the identifiable information therein.
Hardware SetupMy main workhorse is the AIRcable Host XR, an extremely powerful USB Bluetooth device that is primarily designed for proximity marketing. It has a 200 mW radio (twice the power of a normal Class 1 device) and a standard RP-SMA antenna connector, which makes it perfect for long range applications. With the Host XR connected to a directional parabolic antenna, I have been able to establish a connection with a moving cell phone at over 350 meters, careful aiming and a more docile target would push that number even higher. As an added bonus, the Host XR uses the very well supported Cambridge Silicon Radio chipset that I mentioned in the previous article, so it will work out of the box on any modern OS as well as support the extended feature sets used in some software.
If there is any downside to the Host XR, it's cost; at $130 USD you need to be pretty into this sort of thing to make the purchase worth it. Luckily, older versions of the Host XR sell on eBay for around $60, which is considerably more reasonable for the experimenter. The older versions of the Host XR (XR1 and XR2) are identical to each other, the XR2 simply has a nicer case and a bit more mass-market friendly blister packaging (the XR1 I purchased from AIRcable literally came in a Ziplock Freezer bag). The current model, XR3, is claimed to be even more powerful than the XR1/2, but I haven't been able to personally verify this.
The parabolic antenna is great for range, but just a tad bit suspicious when sitting in the food court at the mall. For general scanning, I go with a 3 dBi "rubber ducky" antenna which makes the whole package easy to conceal in a standard laptop case. If I'm going to be scanning from a location where the hardware won't be visible, I bump the antenna up to a 9 dBi (the 9 dBi is about 16 inches long, and tends to get some glances) which pushes the range up to nearly 300 meters. Even with a tiny 3 dBi omni, the Host XR can pickup devices at around 250 meters, which is still twice the "maximum" range for Bluetooth.
That being said, there is no technical reason you couldn't use your machine's internal Bluetooth hardware or a cheap USB adapter. Any Bluetooth device is capable of scanning, the issue is one of power. Using low-power hardware is fine for testing and getting a hang of the software, but with a range of 10 meters (or less), you aren't likely to get many results outside of the devices sitting on your own desk.
Discoverable Device ScanningDiscovery scanning is the most common and effective way of finding information on Bluetooth devices. Fundamentally, it's based on the same process that two Bluetooth devices use when attempting to make a legitimate connection, such as pairing a phone to a headset. Before Bluetooth devices can connect to each other, at least one of them has to publicly announce their presence; discovery scanning exploits that fact to collect information anonymously.
One of the interesting things about Bluetooth discovery scanning versus something like WiFi scanning is that there are three distinct steps required to collect all of the pertinent data for each device. The first step is to command the hardware to scan all available channels for discoverable devices and return their MAC addresses. Once the list of MAC addresses is gathered, each device will be queried individually to determine the device's human friendly name. With the device's MAC and name recorded, the system can then use Service Discovery Protocol to find out what high-level services the target device offers.
It's worth noting that only the first step, getting the MAC address, is technically required to establish a connection with the remote device. Since each subsequent step takes a few seconds to complete, it is often desirable to forgo the third and even second steps for the sake of time and accuracy. This is especially true when dealing with moving targets or when working with short range hardware; if there are 10 devices reported during the initial scan, and it takes 1-2 seconds to complete the second and third steps, the time required to gather all of the information for each device quickly compounds to the point that a particular device may have moved out of range since it's initial discovery.
The following are tools for Bluetooth discovery scanning which are worthy of your attention if you plan on doing experimentation of your own. There are a number of other tools available, but they tend to be only lightly featured or in some cases abandoned.
btscannerArguably the best known Bluetooth scanner, btscanner  is even included in the package repositories of many Linux distributions. btscanner is very easy to use and collects an incredible amount of information without needing to pair with the target devices. The information is presented in a ncurses-based user interface that was clearly inspired by Kismet; something unique among Bluetooth scanners.
If there is any downside to btscanner, it's that it hasn't seen an update since 2004 and it's age is starting to show. There are a number of rather annoying UI bugs, and the more advanced features present in the new breed of scanners is notably absent. An option to compile and run btscanner without ncurses is also sorely missed; and could be a serious problem depending on your requirements.
Stalled development aside, this legacy tool is still useful for the occasional quick scan and the fact that many distributions include it certainly helps keep it popular.
SpoofToophSpoofTooph  is not a Bluetooth scanner in the strictest sense; while scanning for Bluetooth devices is one of it's core features (and something it does very well), it's primarily designed to spoof or clone Bluetooth devices.
SpoofTooph first scans the area to locate devices which are in discoverable mode, then allows you to select one of them to spoof. Spoofing a Bluetooth device can be used for a number of purposes, such as attempting to circumvent applications which use a Bluetooth device (such as a mobile phone) as a security token. You can also load your Bluetooth device up with completely fictitious information, which could be used to make legitimate Bluetooth communication more difficult.
It's more advanced functions aside, SpoofTooph is an excellent scanner as it presents results in a very intuitive and easy to read format in the terminal with minimal configuration or interaction from the user. All discovered devices can be logged to a plain text file, and SpoofTooph even includes a log reader mode where it can reload a previous scan's results for later review and cloning.
Harald ScanHarald Scan  is a modern Bluetooth scanner written in Python, perhaps best known for it's ability to determine device manufacturer via the largest known Bluetooth MAC address vendor list.
Harald Scan offers a number of very nice features not found in other scanners, such as the ability to update it's MAC vendor list via the Internet, optional service scanning, and metrics showing how many devices have been discovered in a given time frame. It's user interface is very simplistic, but gets the point across. Devices are logged in an easily parsed XML format which is ideal for exporting the data into other applications or formats.
Active development, cross-platform support, and unique features make Harald Scan an excellent tool for general Bluetooth scanning.
BluelogAlright, full disclosure time, Bluelog  has been my personal project for over a year now. I wrote Bluelog because all of the Bluetooth scanners I found seemed to be designed around the same basic idea: that you wanted to stare at a display of all the devices being discovered in real time, and maybe save a log at the end of the scan. But what I really wanted to do was scan over a long period of time and log directly to file without having to monitor the software or interact with it in any way.
Because of my rather specific goals, Bluelog is unique among Bluetooth scanners for a number of reasons. The major difference between Bluelog and other scanners is that it has no user interface to speak of, just some boilerplate and status information as it starts up. Though enabling verbose mode will let you see devices as they are discovered, Bluelog's primary method of output is real-time logging to file. By saving results in real-time, you can kill Bluelog at any time and be sure all of your data has been recorded. Bluelog is also (to my knowledge) the only Bluetooth scanner to feature a daemon mode, where it can drop into the background and log discovered devices to file indefinitely.
Probably it's most unique feature is Bluelog Live, a special mode where discovered devices are displayed via a constantly updating web page, with each device's pertinent information laid out in plain English. Inspired by the infamous "Wall of Sheep" display, the goal of Bluelog Live is to raise public awareness of the implications of discoverable Bluetooth devices. Running at a hacker convention or other public gathering, Bluelog Live should provide viewers with an eye opening look at the amount of identifying information they are broadcasting.
Non-Discoverable Device ScanningScanning devices in discoverable mode is easy enough, but what if you wanted to find devices that weren't actively transmitting their presence? While far from ideal, it is possible to scan for non-discoverable Bluetooth devices by exploiting a core concept in the Bluetooth protocol: even if a device is not in discoverable mode, it still has to answer to requests for information so that it can communicate with legitimate peers.
Instead of listening for broadcasting devices, non-discoverable scanning queries individual devices by MAC address for their configuration information (such as the "friendly" device name). The target device is obligated to respond to such requests, regardless of discovery settings and without so much as a prompt on the target device's screen. The problem is, this only works if we already know the MAC address of the target device; which at this point we don't. Since we don't know the MAC address of the target, and it isn't broadcasting, the only remaining way to find the MAC is by brute-forcing it.
To brute-force a Bluetooth MAC, the software will sequentially step through a pre-defined range of MAC addresses, pausing on each one to perform a device inquiry. When and if the scanner receives a response, it can log that MAC and associated information to file. The process can be sped up considerably by using multiple Bluetooth adapters to parallelize the operation, and the MAC range can be narrowed a bit if you know the manufacturer's OUI. Still, given the sheer number of possible MAC addresses and the fact that it takes a few seconds for the device inquiry (whether there is a response or not) to complete, brute force scanning is a very daunting task.
Realistically, scanning for non-discoverable devices is only practical if you are targeting a single device who's manufacturer and make you have already established visually. Even with enough information to narrow your MAC range and multiple adapters working the queue, it's going to take a very long time to complete a single pass. The scope of this method is exceptionally limiting, but if you are targeting something like a Bluetooth enabled desk telephone in an office building, it's possible.
If you want to experiment with brute force Bluetooth scanning, the best tool available is the one that introduced the concept in 2003, RedFang . While it's rather simplistic and hasn't seen development in quite some time, it's still the most reliable tool for this type of device scanning.
Real World ResultsWhile doing some research on past Bluetooth security talks and demonstrations, I found a very interesting white paper put out in May of 2006 by F-Secure and Secure Network entitled "Going Around with Bluetooth in Full Safety" . The paper details how the group constructed a mobile Bluetooth scanning rig disguised as a pull-along piece of luggage and wandered around Milan with it during Infosecurity 2006. After 7 days of scanning at various high-traffic areas, the team recorded 1,405 devices; which at the time made something of a stir and grabbed a few headlines in various tech publications.
The paper goes on to say that their scanning rig never attempted to connect to any of the discovered devices, and that simply being discoverable does not necessarily mean the device is exploitable. The exercise was merely to gauge the proliferation of Bluetooth devices, specifically, those left in the ill advised discoverable mode. The paper concludes that the number of discoverable devices in the field is already high, and that as smartphones become the norm in the near future, this number is only going to go up.
Reading this article 5 years after it's publication, I couldn't help but wonder how the situation may have changed. The authors correctly predicted the era of the smartphone, but had manufacturer's wised up about Bluetooth security?
With this in mind, I setup my own similar experiment which I deemed "Operation Street Sweep" . In my version, I installed Bluelog on an OpenWRT router and connected it to one of my AIRcable Host XR Bluetooth adapters; I then placed the rig within a hundred meters or so of a fairly busy intersection and shopping plaza. After only 5 days of scanning, my setup had recorded a staggering 2,596 unique Bluetooth devices.
I was completely blown away; first by how many devices I was able to record, but also how much information I could glean from them. Taking a close look at the records showed there were all kinds of identifying information hidden within the MAC addresses and device names. For example, Garmin in-car GPS units have their serial numbers in their device names, and the iPhone conveniently broadcasts the owner's full name. I also saw a number of devices who's owners had made the name their online handle; a quick search on Google, and I was well on my way to identifying the individual.
What's more, if we allow devices to be logged multiple times over a long duration scan, the timestamps can be compared each day and a timetable could be constructed for each individual MAC. This rough schedule, paired with the unique identifying information for a specific device, would allow an attacker to get a good idea as to where the target is at a particular time. In it's most basic form, this technique could be used to determine when a target is away from their home.
ConclusionExperiments like this, performed with open source software and readily available consumer hardware, show just how much information is being beamed out for anyone who cares to listen. With thousands of devices eagerly announcing their presence and identifying their owners, even a very low success rate on attacks could be a real threat. Even if we assume that only 1% of discoverable Bluetooth devices are vulnerable to attack (through either social engineering or exploitable implementations), there are enough targets that it's still feasible to hit a few devices a day successfully.
Hopefully this article has given you enough information on the ideal hardware and available software to launch your own research. Keep in mind that the results I obtained in my experiment were not due to some magic combination of hardware, software, and location; I believe these results to be representative of average Bluetooth device density, and should be easily repeatable. I've also performed numerous scans at public locations such as malls, movie theaters, and restaurants. I've found that malls are very good places to conduct scans, especially at peak shopping times like the holidays. In a well populated mall, I've had no problem picking up upwards of 100 unique devices per hour, even with a standard Bluetooth adapter.
In fact, I'm willing to bet that in more densely populated areas my results would be put to shame. I'd be very interested in hearing from anyone who conducts similar experiments using the methods and software mentioned in this article; I've been thinking about a distributed effort to catalog and map Bluetooth devices and their relative density (like WiGLE  for Bluetooth), and would love to compare notes.