From the course: Threat Modeling: Information Disclosure in Depth

Unlock the full course today

Join today to access over 22,600 courses taught by industry experts or purchase this course individually.

AI and machine learning

AI and machine learning

From the course: Threat Modeling: Information Disclosure in Depth

Start my 1-month free trial

AI and machine learning

- [Instructor] There's a lot of information that can be disclosed about a machine learning system. It's tempting to say we're still learning. But we can picture a system. At some level, your training data is under your control. If you disclose information about where those inputs are coming from or how you filter, an attacker might use that to design their attacks. Machine learning systems are hard to tune and so if you have a good model, an adversary may want a copy of it. Knowing how your ML system works makes it easier to design at half data. As we talk about this, let's take a quick dive into the area of security by obscurity. You may have heard, correctly, that security by obscurity is a bad idea and that's right. We intuitively know that the little rock that has your house key in it, isn't as good as a key safe. The rock depends on your attacker not noticing, the key safe has a combination that controls access. So…

Contents