have a look at the link article

https://www.computer.org/csdl/proceedings/big-data/2016/9005/00/07841037.pdf

Deeb Learrning in big data analytics

One of the biggest threats today in industry is we are generating more data than we can process. Most processing takes a lot of time (deep learning) and only a handful of tools that are capable of doing it (Spark + machine learning has seen training time reduction in up to 80%). Also, productionizing a solution that can do even faster and also scalable would be a great research project. In the end, security comes in how your processed data is being consumed or how your data is accessed to be processed. I would suggest you to research tools out there and see what are their challenges. That should help you start. Hope this helps.

Here is a good paper.

http://homepages.cs.ncl.ac.uk/paolo.missier/doc/p56-zaharia.pdf

Thanks you for suggestion Nagdev Amruthnath

Thanks Layth Dhafer Shukur Al-Khafaji

Thanks Ajit kumar Roy