GSOC18: Reverse Engineering AirPlay for VLCa read
I discovered Google Summer of Code quite late. Scrambling together my applications made for a hectic weekend. I had a goal in mind but, as a contingency, used up all three proposals. A few days later an email pinged into my inbox from a familiar name and I was welcomed aboard.
I was bringing AirPlay support to VLC.
In VLC 3.0 there was a considerable effort made that paved the way for my project. Support for Chromecast, a competing protocol, was added along with bonjour monitoring. The next iteration of this now that the foundations were laid was supporting an additional protocol. In addition to this, I wanted to try and implement streaming the other way: to an instance of VLC. This would mean first reverse engineering the protocol as well as implementating AirPlay’s pairing process. From there it is “easy” as you only need to host the media file and send the URL to the device.
The first step was documenting the protocol which meant a good few days in Wireshark. The result was compiled in the project wiki and served as a reference for the rest of the project. This was continually refined during the course of the summer as new information replaced the old, or more detailed data was uncovered.
The next step was pairing with the TV.
When we set out neither myself nor my mentor anticipated how cryptographically heavy this project would end up being. The reverse engineering was fairly simple with a few projects that I could learn from that had partially implemented it. The main problem was translating these high-level implementations into low-level manipulation of cryptographic primitives. Some of the libraries we used had difficult to reason with APIs and counterintuitive quirks that made getting it working hard. With no indication of progress from the Apple TV it was hard to determine whether any progress made was even in the right direction.
So after a long couple of weeks of entirely unforeseen work, the first signs of progress started to appear. The Apple TV would respond to the pairing requests and we were able to send arbitrary URIs to be displayed on the screen.
From there we, like in the Chromecast module, had to simulate a sub stream and create a web server to serve the playlist files for the Apple TV.
The result of this project can be found at the link below.
The scope of the project was reduced as we realized just how much work the pairing process would take. As such, I have not managed to implement casting to VLC over AirPlay, only from VLC over AirPlay.
However, I have learned an inconceivable amount the last few months and the delays were not due to lack of progress. Any work that didn’t result in direct progress was instead contributing towards my education. I have read multiple books on cryptography, code refactoring, concurrency and threading, and POSIX. Beyond that, I have learned a great deal about the C programming language (and c++ too), memory management, networking, sockets, file descriptors, multithreading, ECC, SRP6, and on and on not to mention the good habits. Although I’m worried I’ll start using Hungarian notation all the time!
I originally applied as a student looking to get into open source and I have achieved all of my goals. My original draw to this project was due to my long history with the software and it feels great to be given the opportunity to give back some of the hours I spent using this software. I plan on continuing work on VLC in the future.
I’d like to thank
- Steve Lhomme, my mentor, for answering my endless questions, dealing with my odd schedule, and keeping me on track
- JB, the rest of the people at VideoLabs, and the rest of the contributors for maintaining such a great project
- funtax for their high-level AirPlay pairing implementation saving me many hours
- cocagne for their implementation of SRP6 from which I based my version
Lastly, my girlfriend and family, with whom I have celebrated the good times and persevered the bad times.