A team at IBM recently presented a new technique through which Ai developers would be able to secure their intellectual copyrights. It was presented at the ACM conference on Computer and Communications Security (ASIACCS) 2018 in Korea.
Through this new advancements, deep learning methods can be protected from attackers. The system is a robust model that would halt stealth right in its tracks. The IBM version through embedding would allow application of watermark on deep learning networks and then its detection and extraction to prove ownership.
The watermark has no impact on the performance of the neural network. The project is in its very early stages and if proved successful, IBM plans to commercialize it.