Chapter 9 – Bypass Machine Learning Malware Detectors
What are the components of generative adversarial networks?
The two main components of a generative adversarial network are the generator and the discriminator.
What is the difference between a generator and a discriminator?
The generator takes latent samples as input. They are randomly generated numbers and they are trained to generate images, while the discriminator is simply a classifier trained with supervised learning techniques to check whether the image is real (1) or fake (0).
How can we make sure that the malware adversarial samples are still valid when we are generating them?
To avoid invalid samples, we can use a Sandbox/Oracle.
Do a bit of research, then briefly explain how to detect adversarial samples
To detect adversarial samples, we can remove the noise by using binary thresholding.
What distinguishes reinforcement learning from deep learning?
Reinforcement learning learns how to maximize a reward function by exploring the actions available from certain states, while deep learning learns from examples it has been shown.
What is the difference between supervised and reinforcement learning?
In supervised learning, given input data X and labels Y we are learning a function f: X → Y that maps X to Y. In reinforcement learning, the agents are getting smarter after number of experiences.
How does an agent learn in reinforcement learning?
In reinforcement learning, the agent learns by interacting with an environment based on a reward function to optimize its performance.
