Jump to content

AceKingSuited

Members
  • Posts

    6
  • Joined

  • Last visited

  • Feedback

    100%

Posts posted by AceKingSuited

  1. 8 hours ago, Naked said:

    Pretty sure he wants to override the mouse path, not just how to instantly click it with mouse.click

     

     

    Yeah, overriding mouse path was exactly what I was looking for.

     

    12 hours ago, BravoTaco said:

    If you just need the position of whatever you are trying to click, and the object implements Entity, you can retrieve the bounding box of said object. Than use your mouse movement logic to move within the bounding box, than interact with it.

    This is what I was referring to with leveraging API, awesome.

     

    12 hours ago, Medusa said:

    Only problem with this would be that the boundingboxes sometimes are too big.

    
    public static void clickObject(MethodProvider api, RS2Object o, boolean rClick) {
    		int x = (int) o.getModel().getBoundingBox(o.getGridX(), o.getGridY(), o.getZ()).getX();
    		int y = (int) o.getModel().getBoundingBox(o.getGridX(), o.getGridY(), o.getZ()).getY();
    		api.getMouse().click(x, y, rClick);
    	}

    That will get the boundingbox of an object and click it, but if it's a ladder or some shit it will misclick a lot

    You probably just saved me a *lot* of debugging time lol, thanks

     

    8 hours ago, Medusa said:

    Shoo. No one wants to hear about your mouse paths.

    Actually I do 😭

  2. I was wanting to implement my own mouse movements generated via machine learning against human data. That being said, I don't want to go manually calculating coords and hitboxes if at all possible.

    Is there any way to override the default mouse movement when entity.interact is called? I'd basically like to just rip out OSBot's default implementation of mouse movements and replace with my own.

    Any advice would be awesome, just trying not to reinvent the wheel and go back to calculating / learning XY coords for everything. Makes more sense to leverage what the API already knows (coords / size of clickable areas, etc)

  3. This thread has brought a lot of confirmation to what I've spent the last few weeks mulling over. Even down to the recording of movements to compare against the client's mouse movement and attempting to see if the client's mouse movement can easily be identified. Always suspected mouse movement was sent to the server, but never came across any physical proof until reading this thread.

    There's a project we'll call "Cat and Mouse" that I've been thought experimenting with for a while to help with this *exact* problem. While part of me has been wanting to keep it to myself I think I'll share here where it feels most relevant and maybe even open source it at some point if it makes sense to. Recording mouse data on a large scale is difficult (consent, incentive, etc) so the point of this project would be to do it in both an ethical and incentivized manner.

    Basically, there are 3 objectives to "Cat and Mouse".

    1. Collect human mouse data (and maybe keyboard or other forms of behavior down the road) via a fun / incentivized game designed to create situations that affect behavior. Some examples would be size of click area, distance of movement, importance of accurate clicks, moving vs fixed targets, whether the movement is rushed, etc
    2. Create a bot that attempts to go undetected playing the game vs that large sample size of human data <---- Mouse
    3. Create a bot detection system that the bot is constantly trying to beat <---- Cat

    With a large sample size of human data and several iterations of the bot detection system (via machine learning?) the bot should be able to get better and better at replicating human inputs (mouse, keyboard, etc).

    I'm sure there's a lot more to it than just what I've described as "solving the problem" and I'm purposely keeping a lot of the more interesting bits to myself, but coming across this thread kind of sparked the need to share the idea :) 

×
×
  • Create New...