How to train a robot to mimic how I browse the web?

A place to discuss the implementation and style of computer programs.

Moderators: phlip, Moderators General, Prelates

jacques01
Posts: 42
Joined: Thu Oct 08, 2015 4:56 am UTC

How to train a robot to mimic how I browse the web?

Postby jacques01 » Sat Jul 23, 2016 2:09 pm UTC

I would like to train a browser robot to work like me, or at least human-like, in its interactions with various websites.

The robot would interact with a website like a human (or like I do). What it can't really be taught is what its payload is, e.g. does the robot check "read" on all my unread emails. I will tell it what the payload is beforehand. What I want to teach it is how to "get there" in a human way.

My idea is that once I train the robot correctly, if someone took analytics on both myself and the robot, they could not reliably distinguish the two (it would be a coin flip, e.g. like the machine generated poetry). Of course, they could use the fingerprint to still know whether it's Jack or Bob, but if the robot was trained in Jack, robot-Jack and real-Jack are just Jack to the outside world / observer.

Here is my approach, for which I'd need at least 2 things (note I'm gathering data on myself, which is 100% legal):

1) Collecting training data.
- How can I get analytics on my web browser usage patterns in a structured and easy to interpret way?
This means things like mouse tracking, time spent viewing a page, scrolling direction/speed, typing speed
Basically, any activities which a robot can accomplish inhumanly fast but I want it to act like a human.
These are also indexed by time as well, e.g. I'm quicker on the browser in the late morning than at 2:00 AM.
Other human things include hovering on ads, getting distracted by content/feeds (I was just going to send a Facebook message but I saw a cool article pop up in my feed). A robot wouldn't do those things unless it's part of the payload.

2) Training the robot.
-I have collected millions of samples of training data for different tasks.
- I index these per website or even per page (e.g. I behave differently when writing an XKCD message than I do when viewing the XKCD forum index).
- This labeled data comes for "free" if I do my analytics / tracking correctly.
- Which machine learning algorithm(s) do I use?
- Is this really just a probabilistic finite state automata ?
- How many classifiers / stochastic generators would I need, e.g. I have one for mouse cursor movement, one for typing speed / rate, one for viewing the page duration, one for scrolling, etc.

What about alternative approaches?

commodorejohn
Posts: 957
Joined: Thu Dec 10, 2009 6:21 pm UTC
Location: Placerville, CA
Contact:

Re: How to train a robot to mimic how I browse the web?

Postby commodorejohn » Sat Jul 23, 2016 5:17 pm UTC

"Now that I've trained a robot to do my web browsing for me, I can finally get some real work done!"

[...]

"...maybe just one more quick look at Reddit..."
"'Legacy code' often differs from its suggested alternative by actually working and scaling."
- Bjarne Stroustrup
www.commodorejohn.com - in case you were wondering, which you probably weren't.


Return to “Coding”

Who is online

Users browsing this forum: No registered users and 13 guests