The paper is here https://github.com/NationalSecurityAgency/fractalrabbit/blob...

Interesting that the model is based on observed animal behavior. This paper appears to be the source for that: "Random walks with preferential relocations to places visited in the past and their application to biology" https://arxiv.org/abs/1403.6069

The animals in question are not rabbits - I thought the rabbit was metaphor anyway - but capuchin monkeys.

I can't get enough of this sentence: The FRACTALRABBIT stochastic mobility simulator creates realistic synthetic sporadic waypoint data sets.

TIL the NSA has a GitHub. I guess it makes since, but still. Huh.

Also home to Ghidra, a very popular free open source reverse engineering tool: https://github.com/NationalSecurityAgency/ghidra

As far as I'm aware, there is only one tool that is better (IDA Pro), which requires a ~3k/year license.

So ... generating plausible test data for pattern/deviation detection of surveillance?

Googling "co-travel algorithms" yields only one result! I would also assume it means surveillance or pursuit.

I understand a trajectory for animal motion, but how can there be trajectory for websites? The state vector is virtual. The space is not meaningfully physical (0). One can move from any IP address to any IP address in microseconds or less. The distance between these movements is virtually identical. How can one confidently predict anything in that space using an animal model?

(0) put to the absolute extreme, there are limits on how fast a computer can move the numbers from one memory cell to another, but that's a different universe from self-powered locomotion such as a rabbit, monkey or human moving in meatspace.

The space could be a graph ("network") and the edges could be e.g. links or some measure based on similarity.

The companion git repo is fascinating. https://github.com/probabilist-us/contact-rabbit

Trying to reverse engineer indirect infection.

This is useful for seeding propaganda and any other virally driven phenomena by modeling gathered exploratory behavior of a user.

This looks fascinating, but I don't have the math to really understand. I can see I'd need linear algebra and some statistics, but what else would be required to understand and work with something like this?

I'm impressed that they managed to shove in some weird Powerpoint-style formatting at the beginning of the readme.

What do you mean? Inserting an image and header text and centering everything with HTML? https://raw.githubusercontent.com/NationalSecurityAgency/fra...