Can you make a Raspberry Pi Security Camera?

This weekend’s geeky challenge: Making a Raspberry Pi Security Camera.

TL;DR – Here’s a basic, practically-ready-to-deploy, open-source Raspberry Pi security camera.

Brief

Mostly because I’m cheap (but partly to reduce the scope and therefore increase chance of success) I added a few constraints:

  • Periodic still image capture is OK (streaming video will be tackled in a future iteration)
  • Securely store images in the cloud (so I can check them when I’m away from home)
  • Be configurable
  • Support the cheap official Raspberry Pi camera modules, particularly the “NoIR” variants which come without the Infrared filter for night-time captures when complemented with an IR light pack

Let’s get started.

Design

After a quick napkin sketch it becomes clear we need to build 4 key parts:

  • A .NET Core console application.
    • This is our entrypoint, and using .NET Core lets us compile to the “linux-arm” runtime target, i.e. Raspbian.
  • A timer
    • To periodically take the photos. I have some ideas for other conditions or events that could trigger the camera to take a snapshot (such as a PIR sensor detecting heat) so I’ve implemented the timer as a “TimerTrigger“, which implements an ITrigger interface. That way it’s nice and extendable.
  • Some code that speaks to the Raspberry Pi camera module
    • Again I’ve written a “CameraImageSource“, which implements the IImageSource interface.
    • I also wrote a TestImageSource which loads an image from the current working directory for testing.
  • A client for the cloud storage API
    • I’ve chosen Google Drive because it offers 15GB free storage, a decent SDK that supports .NET Standard (1.3) and has good documentation.
    • As with the other components, this is written as an implementation of the “IDataStore” interface, so it’s easy to expand to other cloud storage providers in the future.

And this is how they all communicate:

Implementation

The code is here:

https://github.com/richstokoe/pispy

Once you’ve cloned that repository, there are some things we need to do before we can run the code.

Cloud Storage Credentials

To use the Google Drive API, you need an OAuth2 client ID and client secret. Follow these instructions to get a client ID and client secret, then hit the Download button in the “Credentials” area of the API console and save the file as “client_secrets.json” in the src/PiSpy/ folder of the repository you just cloned.

This file will be copied to the output directory when the project is built, per MSBuild instructions in the PiSpy.csproj project file.

Configure the Raspberry Pi Security Camera

If you want to change the default options, open up the appSettings.json file in the project.

Timer interval

You can change the interval between camera shots by modifying the Triggers:TimerTrigger:Interval setting. This is in milliseconds (seconds / 1000 so 60 seconds = 60000).

The default is 180 seconds (3 minutes).

Camera output directory

You can change the output directory by modifying the CameraModule:OutputPath setting.

The default assumes you will copy the console app to /home/pi/pispy, if you want to copy it somewhere else, change the path accordingly, otherwise you’ll see an error from the “mmal” process.

Raspbian with a Desktop

For now, we need an operating system on the Pi with an interactive desktop to complete the Google OAuth2 authorization flow.

Follow the steps in my last blog to set up a Pi with Raspbian and the .NET core runtime – but with one minor difference: grab Raspbian Stretch, NOT Raspbian Stretch Lite. This gives you a Desktop – that’s important for when the Google authorization flow pops open a browser window to enter your Google account details.

(In a future iteration we’ll add a Kestrel HTTP endpoint to the service to negate the need for the non-Lite version of Raspbian with a desktop).

You can either plug in a HDMI-capable screen and keyboard, or you can enable VNC by running:

sudo raspi-config

at the command line (or via SSH) and enabling ‘VNC’ under the ‘Interfacing options’ menu.

I also found I had to set Chromium as my default browser for the authorization flow to work properly – to do that navigate to chrome://settings, choose “Set Chromium as my default browser” and then restart the pi by running:

sudo reboot

Deploy and run

As in my last blog, build the project targeting the linux-arm runtime:

dotnet publish -r linux-arm

and copy the the bin/Debug/netcoreapp2.0/linux-arm/publish folder via FTP to the pi. I copied it to /home/pi/pispy.

We’re going to need access to Raspbian’s PIXEL desktop in a moment, so connect to the Pi’s VNC server by following the instructions here. Once you’re in, open a Terminal and navigate to the folder you copied the /publish folder to. Create a new folder within it called stills, which is where the CameraImageSource code will write the pictures to before handing us a Stream.

cd /home/pi/pispy
mkdir stills

(I’m cheating a bit by simply running the “raspistill” executable that comes with Raspbian to take the pictures. More info here.)

Next, start the service by running:

dotnet /home/pi/pispy/PiSpy.dll

After the time specified in the Triggers:TimerTrigger:Interval appSetting elapses, a photo will be taken and then the GoogleDriveDataStore will trigger the authorization flow (this only needs to happen once). Once you’ve logged in future photos will stream up to the Google Drive.

Next steps

Pull requests are not only welcome, they are encouraged.

A good place to start: I haven’t had much time this holiday season to debug (same excuse for shoddy blogging 🙂 ), but there appears to be an async bug in the TimerTrigger whereby the subscribed actions are invoked on a separate thread and the Timer is restarted even though the CameraImageSource still hasn’t finished taking its picture.

And don’t forget to turn off your LEDs:

I hope you’ve enjoyed this slightly rushed guide to building a Raspberry Pi security camera!

Running an ASP.NET Core 2.0 app on Raspbian Stretch Linux on a Raspberry Pi with HTTPS

Today’s challenge: Serve a public API over HTTPS from a Raspberry Pi. I’ll follow up with an article about containerising the app and running it from docker on the Pi.

This article is a fast-paced guide to getting started without stopping to dwell on the details.

Some details will be mopped up in the last section for those wanting to know more.

Prepare the Pi

On your development machine (you won’t need to interact with the Pi directly, so no 2nd keyboard or monitor required):

  • Grab an 8Gb or larger microSD card and use Etcher to flash the Raspbian Lite image to the card.
  • Create an empty file in the root of the microSD card called “ssh”, with no extension, this will enable ssh on Raspbian.
  • Put the microSD card in the Pi, plug in a network cable and then plug in the power
  • After a minute attempt to ping the pi:
ping raspberrypi
  • If you get a response, ssh into the pi from a Bash shell (on Windows, you can use the bash shell that comes with Git):
ssh pi@raspberrypi
  • The “pi@” means you’re logging in as the “pi” user account. The default password is “raspberry”.

Installing the .NET Core prerequisites

This bit looks hard but is quite easy (thanks Dave).

At the ssh prompt, install the .NET dependencies by running each of the following commands (you can copy and paste these commands straight into bash):

sudo apt-get install curl libunwind8 gettext
curl -sSL -o dotnet.tar.gz https://dotnetcli.blob.core.windows.net/dotnet/Runtime/release/2.0.0/dotnet-runtime-latest-linux-arm.tar.gz
sudo mkdir -p /opt/dotnet && sudo tar zxf dotnet.tar.gz -C /opt/dotnet
sudo ln -s /opt/dotnet/dotnet /usr/local/bin

you can check if the .NET Core runtime has been installed by running:

dotnet --help

You should now see some the runtime command line options.

Configuring the App

If you don’t already have an ASP.NET Core app:

  • Download the SDK for your platform (Mac, Linux, Windows) from https://dot.net and install it
  • Make a new directory and navigate to that directory at the command line.
  • Run:
dotnet new react

By default, a new ASP.NET Core application will be set to only listen to requests only from the TCP/IP loopback address (127.0.0.1 or localhost), so:

Open up Program.cs in your ASP.NET Core app and change the BuildWebHost method to add the following:

public static IWebHost BuildWebHost(string[] args) =>
    WebHost.CreateDefaultBuilder(args)
        .UseKestrel(options => {
            options.Listen(System.Net.IPAddress.Any,
                5000,
                listenOptions => {
                    listenOptions.UseHttps("raspberrypi.pfx", "<your-password>");
                }
            );
        })
        .UseStartup<Startup>()
        .Build();

This means that when the app runs on the Pi it will accept requests to port 5000 from external clients.

Please note, Kestrel is not a supported edge server, it is designed to run behind a reverse proxy such as nginx, Apache HTTP Server or Microsoft IIS when exposed to the outside world. Read and understand this before you open up your Pi to the big bad interwebs.

Deploying the App to the Pi

I’m not offering DevOps perfection here, I’m afraid we’re just going to FTP the app across to the pi. But first we need to compile the app so it works with the Raspberry Pi’s low-power ARM processor.

On your development machine, drop to the command line, navigate to your project directory and publish your app so it works on Raspbian by executing the following command:

dotnet publish -r linux-arm

This creates a bin/Debug/netcoreapp2.0/linux-arm/publish directory that contains the binaries for your ASP.NET Core app.

Grab your favourite FTP client. If you don’t have one, FileZilla will do the trick.

Connect your FTP client to your Pi by entering the following details:

Host: raspberrypi
Username: pi
Password: raspberry
Port: 22

The FTP client should show you the directory structure on the Pi. Copy the contents of your linux-arm/publish directory to any path on the pi (I chose /home/pi/piservice/) using the FTP client.

HTTPS

We’re going to use a self-signed certificate to show HTTPS is possible. In a real-world scenario you’d sign a relatively short-lived RSA keypair with a certificate signed by a trusted root cert (and also, you’d probably not use a Raspberry Pi and publicly-exposed Kestrel Web Server to run your services, but hey ho).

SSH into your Pi again and run the following command to create a public and private key pair that will be valid for a year.

openssl req -x509 -newkey rsa:4096 -keyout key.pem -out cert.pem -days 365

Next, generate a .PFX file from the two generated PEM files (thanks to Pete S Kelly):

openssl pkcs12 -export -out raspberrypi.pfx -inkey key.pem -in cert.pem

This creates the “raspberrypi.pfx” file that your ASP.NET Core app now refers to in Program.cs.

Running the app

At the SSH prompt, tell the .NET Core runtime on teh pito start your app:

dotnet <yourappname>.dll

<yourappname> will be the name of the project you created if you used Visual Studio, or the name of directory you created that you ran dotnet new in. Typically this will be something like HelloWorld.dll or Acme.Web.dll, etc.

If all has been successful you’ll see the following echoed in your SSH session on the Pi:

pi@raspberrypi:~/piservice $ dotnet src.dll
Hosting environment: Production
Content root path: /home/pi/piservice
Now listening on: https://0.0.0.0:5000
Application started. Press Ctrl+C to shut down.

(My emphasis). From your development machine you should now be able to begin making requests to your app hosted on the Pi, https://raspberrypi:5000/

You will get errors about the certificate being untrusted, this is expected as your development machine has no reason to trust the little $30 computer, but you can skip past them – or read the final section of Peter Kelly’s article to learn how to trust the Pi’s self-signed certificate.

Making it Public

This is where things turn a bit vague as it’s up to you how you set up your network.

At the most basic level you need to tell your router to send traffic to port 5000 on your Pi. This usually involves adding a Port-Forwarding Rule. You’ll need to know your Pi’s IP address to set up the rule, so it makes sense to either give the Pi a static DHCP lease. Please refer to your router’s user guide for specific information.

To call your Pi from the outside world you’ll need your router’s public IP address – of course it’s best if this is static, ask your ISP if this is an option – and then you can set up a domain name to point to this IP address.

You can get a proper SSL certificate for your domain name from LetsEncrypt, or any other certificate provider.

Note: You’ll get SSL certificate errors when using a self-signed certificate, and browsers may stop allowing access to sites where the certificate doesn’t match the public domain name.

Follow Up

I’ll follow up soon with an article on containerising the ASP.NET Core app and running it on Docker on the Pi.

 

Oracle Appeals Google’s Java Win. What’s next for Dalvik?

Larry Ellison points finger at Larry Page. Shakes it. Offers cookie. Redundantly.

This week, Oracle appealed Google’s win at the patent trial over the Java application programming interface (API) being used in the Android operating system.

In a new document published this week (PDF), Oracle’s lawyers start by introducing a character called Ann Droid. And then tell a story about Ann ripping off Harry Potter: The Order of the Phoenix. Quite strange, but hilarious to think of a serious legal department putting this into such an important document:

 

Ann Droid wants to publish a bestseller. So she sits down with an advance copy of Harry Potter and the Order of the Phoenix—the fifth book—and proceeds to transcribe.

She verbatim copies all the chapter titles—from Chapter 1 (“Dudley Demented”) to Chapter 38 (“The Second War Begins”). She copies verbatim the topic sentences of each paragraph, starting from the first (highly descriptive) one and continuing, in order, to the last, simple one (“Harry nodded.”). She then paraphrases the rest of each paragraph. She rushes the competing version to press before the original under the title: Ann Droid’s Harry Potter 5.0. The knockoff flies off the shelves. J.K. Rowling sues for copyright infringement.

Ann’s defenses: “But I wrote most of the words from scratch. Besides, this was fair use, because I copied only the portions necessary to tap into the Harry Potter fan base.”

Obviously, the defenses would fail.

Defendant Google Inc. has copied a blockbuster literary work just as surely, and as improperly, as Ann Droid—and has offered the same defenses.

Continue Reading “Oracle Appeals Google’s Java Win. What’s next for Dalvik?”