deep learning is increasingly used as a building block of security systems.
Unfortunately, neural networks are hard to interpret and typically opaque to
the practitioner. The machine learning community has started to address this
problem by developing methods for explaining the predict