Paper: Probabilistic Naming of Functions in Stripped binaries
Aim: Name functions in a stripped binary
The paper itself was very difficult to follow. First, my limited understanding of AI didn't help and the general flow was not clear. Right off the start the paper state it is about naming functions. This is clear and why its needed is also clear. They want function naming inorder to assist reverse engineering processes by naming unknown function by their behavior.
Not really sure the consequence it may hold if it labeled it wrong (kind of misleading) but still interesting and useful. They analyze the binary and draw CFG and using trained knowledge it could generate some vector feature which could be placed in another graph to generate the symbol.
This is the part where I got lost. The diverse approaches the paper took was interesting but it rarely justified its choice in implementing something in some way.
Results as always was usually positive (I mean who would post anything negative about their work). However it did seem the recall was much higher than the precision. This was interesting as detailed data was not given so I understand that the score is good but I don't understand why its good. Maybe its not given because it was a 10 fold analysis?
==================================================
Random stuff:
Now I kinda know how Unity works. The youtube series does present a guide to making a game but I think I might try myself. I really don't know what I could make though. I want to make a mini game, very simple.
Maybe I could incorporate AI? Or just make a generic platformer. Not too sure would have to think later.