It really is. From where I'm standing, it seems like just about everything I can think of (and everything I can't think of) has been done already. Cybersecurity is a field that is vast, growing, and relatively immature, yet my knowledge of it is far more immature. Writing a simple script or program is almost completely out of the question. What problems, if any would they solve? I certainly can't think of any that can be solved with rigid code. I can't help but return to machine learning, no matter how long I ponder my idea for a final product. Perhaps it's because I tend to focus on the new and emerging stuff. Perhaps it's because I'm already familiar with it to some degree. Regardless, any project attempting to integrate ML into a cybersecurity framework will be a necessarily gargantuan task. At the moment, the only way I can see ML being useful is with respect to threat context. The job of security analysts in any company is often to synthesize information from blogs, news, and professional experience to evaluate whether likely threats detected by their system warrant further investigation. That context is something that is oftentimes not quantifiable, and seems like the perfect problem to apply neural networks to. This is purely conceptual, though, and I really need to meet with my mentor again to discuss this new idea.