I'm not sure if this has been brought to light before, but it's possible to make an orbwalker with champion targeting without modifying process memory.
I implemented this with computer vision and deep learning to classify enemy champions on screen in real time.
I thought processing speed would make this impossible, but my CPU isn't even that strong - the screen capture that feeds into my model runs at less than 2 FPS and the results are still pretty **** accurate.
Just thought i'd throw this out there, if anyone wants to try this method..






