An Introduction to Secure Machine Learning
Machine Learning has created a seismic shift in how we think about software engineering and building things in general. Instead of asking how we should solve things ourselves, we dream of being able to just point at the data and let probabilistic systems make an objectively accurate decision for us.
Lately, there has been a growing conversation about the various risks inherent in these sorts of models and how we should be mitigating against them. The field of secure machine learning attempts to address this problem. The idea in its most basic form is to act proactively by putting yourself in the mindset of a criminal and ‘hacking’ into your own machine learning systems to identify problems, weak spots, or potential back doors before an actual criminal can do the same. After reading this book, you will understand the fundamentals and how you can dive deeper into the field of secure machine learning.