idea: use machine learning to distinguish between "normal traffic" / polling attempts and exploit attempts

[ghost]

Have some ideas for this, although not sure how well it will work. Any ciphertext will completely break it.