Setting Up The Nuand BladeRF On Mac OS X for Bluetooth Low Energy (BLE) Hacking

(Cross-posted to peerlyst.com)

This is a quick guide to getting the Nuand BladeRF software defined radio (SDR) up and running on Mac OS X. The end goal is to have an environment that can be used to perform basic Bluetooth Low Energy (BLE) hacking. The initial focus will be getting it set up to perform BLE scanning and advertising.

A note on formatting: I’ve bolded commands that should be entered in at the command prompt. Any other code formatted content is output from running the commands, shown for reference.

Here’s the steps needed:

1. Install the dependencies, clone the repo, compile the software:

brew install libusb pkgconfig cmake libtecla
git clone --recursive https://github.com/Nuand/bladeRF.git
cd bladeRF/host
mkdir build ; cd build
cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/opt/local ..
make
sudo make install

2. Make sure the device is recognized by the system:

$ bladeRF-cli -p
  Description:    Nuand bladeRF
  Backend:        libusb
  Serial:         7ddf0461b8c6cb36ffe8358189bc5d1d
  USB Bus:        20
  USB Address:    6

3. Download the most recent FPGA image

  • Make sure you get the correct version for your FPGA size. 
  • You can get the size by running: `bladeRF-cli -e info`
  • Look for a line similar to: `FPGA size:                40 KLE`

4. Setup the FPGA auto-load up so you don’t have to do this manually every time you reconnect the SDR:

  • Create the Nuand directory: `mkdir -p $HOME/.Nuand/bladeRF/`
  • Move the FPGA image to the Nuand directory: `mv $HOME/Downloads/hostedx40-latest.rbf $HOME/.Nuand/bladeRF/hostedx40.rbf`

5. Run the CLI, and you will likely see a warning message, followed by the bladeRF command prompt:

$ bladeRF-cli

table not found. Manual gain control will be used instead.
[INFO @ /<redacted>/bladeRF/host/libraries/libbladeRF/src/board/bladerf1/bladerf1.c:1706] To enable AGC, see "Generating a DC offset table" at https://github.com/Nuand/bladeRF/wiki/DC-offset-and-IQ-Imbalance-Correction
bladeRF>

6. Setup the AGC. This can be done either manually each time, or a table can be created:

  1. To set the AGC Manually:

    1. Set the frequency you wish to use. Since BLE uses 2.4 GHz, we’ll want to use that. In the example below, we set the frequencies to BLE Frequency Channel 37, which is used for BLE advertising. Channel 37 is centered at 2402 MHz, so we can set transmit to 2401 MHz, and receive to 2402 MHz:

    bladeRF> set frequency tx 2401M

      TX1 Frequency: 2401000001 Hz (Range: [237500000, 3800000000])

    bladeRF> set frequency rx 2402M

      RX1 Frequency: 2401999999 Hz (Range: [237500000, 3800000000])

    2. Set the gain (we set it to 8 dB here. There’s nothing magic about that, I picked it on a whim):

    bladeRF> set rxvga2 8
      Note: This change will not be visible until the channel is enabled.
      Setting RX1 rxvga2 gain to 8 dB
        rxvga2:    6 dB (Range: [0, 30])

    3. Calibrate the LMS module:

    bladeRF> cal lms
      LPF tuning module: 23

      TX LPF I filter: 27
      TX LPF Q filter: 39

      RX LPF I filter: 30
      RX LPF Q filter: 30

      RX VGA2 DC reference module: 25
      RX VGA2 stage 1, I channel: 37
      RX VGA2 stage 1, Q channel: 45
      RX VGA2 stage 2, I channel: 30
      RX VGA2 stage 2, Q channel: 33

    4. Run the DC offset calibration routines (look for `Error` values to be near 0:

    bladeRF> cal dc rx

    RX DC I: Value =   352, Error =  0.456
    RX DC Q: Value =   640, Error =  0.265

    bladeRF> cal dc tx

    TX DC I: Value =  -208, Error =  0.347
    TX DC Q: Value =   160, Error =  0.349

  2. To generate a table:

    1. Run the cal command. This will take some time, and should be done after the bladeRF has been running a bit so it is properly warmed up:

    bladeRF> cal table dc rx
    Calibrated @  237500000 Hz: I= 384 (Error: 3.63), Q= 224 (Error: 0.95)      DC-LUT: Max (I=-46, Q=-38) Mid (I=  7, Q= 10) Min
    Calibrated @  247500000 Hz: I= 384 (Error: 2.44), Q= 224 (Error: 1.81)      DC-LUT: Max (I=-43, Q=-39) Mid (I=  7, Q= 10) Min
    Calibrated @  257500000 Hz: I= 384 (Error: 0.81), Q= 224 (Error: 0.30)      DC-LUT: Max (I=-41, Q=-39) Mid (I=  8, Q= 10) Min
    ...
    Done.

    2. Once complete, check the current directory on the file system. There should be a file named `<serial #>_dc_rx.tbl`.

  • For example: Using the device shown in the info command above, the file name would be `7ddf0461b8c6cb36ffe8358189bc5d1d_dc_rx.tbl`

    3. Move the file to the Nuand directory: mv 7ddf0461b8c6cb36ffe8358189bc5d1d_dc_rx.tbl $HOME/.Nuand/bladeRF

That’s all there is to it. Now when you run `bladeRF-cli`, the FPGA will be auto-loaded, and the AGC table will be loaded as well, leaving you ready to rock. In a future post, I’ll talk about where to go next to start listening to beaconing BLE devices near you.

quick and dirty web screenshot tool

In a recent test, I had a need to throw a list of websites at a tool to get a screenshot of each site, so I could figure out what they were quickly.

There are several tools out there that can handle this sort of thing, but none of them were “quick and dirty”, which is what I really wanted: I didn’t want to spend a bunch of time configuring things, and I didn’t want to send a ton of command line arguments, I just wanted to dump a text file of URLs (obtained via nmap scanning) and point a script at it and have it go grab a snap of whatever was running at the URL.

Since I couldn’t find something like that, I just wrote my own. It turns out that all it takes is around 50 lines of JavaScript.

If you ever have a similar need, feel free to check out https://github.com/rossja/url2screen and see if it does what you want. It worked perfectly for me =)

Dockerized Vault with Consul Backend

I recently wanted to get a working vault instance running. Since my goal is to sharpen my devops skillset, I had some specific features I wanted to make sure I had nailed down:

  • Run vault in server mode, rather than dev
  • Run it in Docker, ideally using docker-compose
  • Have other containers access the vault for secrets
  • Use a persistent storage mechanism for the backend
    • I didn’t want to use files for this

To meet the storage requirements, after looking into it a little bit it became apparent that Consul was probably the right move (as of this post, it’s the only database engine that’s officially supported by Hashicorp for Vault — which makes sense since they make both products).

It turns out that getting Vault and Consul up and running – and talking to each other – is not as straightforward as it could be. I found lots of examples, but most of them did not work anymore, and all of them used version 2 of the docker-compose spec. After much trial and error and a ton of failure, I finally got things running successfully. If you want to try it out, you can check it out at https://github.com/rossja/docker-vault-consul/

One thing to note: as configured in the repo the Vault container talks to Consul over HTTP. Since the secret is encrypted inside vault, no sensitive data is being passed in plaintext over the network (in theory), but it’s still not the best-practice to use HTTP here.

The problem is that there’s a bit of a chicken/egg problem with using HTTPS… specifically: you need to securely get certificates and privatekeys into the Vault and Consul containers in order for TLS to work properly. But doing that means either:

1) checking the certs into a repo, or
2) passing them into the container via something like a volume or ENV var.

Neither of these are ideal.

I’ll probably ultimately end up opting for passing them in via a bootstrap volume of some kind, but am still thinking over how best to handle this.