I wrote a PyTorch implementation of Adversarial patch here.
It creates visible adversarial noise that does not alter the semantic meaning of a digital image.
I have been participating in root-me CTFs and war games. Root Me is a platform for everyone to test and improve knowledge in computer security and hacking, and I would definitely recommend checking it out. I have also been working my way through CryptoPals challenges. You can find my solutions here.
Security & crypto resources
For our work on website fingerprinting, I needed to keep fine-grained information about a browsing session. I couldn't find a suitable tool so I created one myself. Although very early in development it did the job.
Numerai is Kaggle-like competition on encrypted data. During the first year of it's launch I tinkered with a few models that were moderately successful. Although the competition
has probably progressed since then I am making available some of the code I used here.
I trained a GAN to learn what a face looks like and then produce its own examples. Read more about that here.
I wrote a small educational guide of the dangers of browsing without HTTPS. Read more about that here.
Decrypt WLAN traffic
An(other) interesting property of the Mandelbrot set
This was taken from my Masters thesis and has always fascinated me.