HN Explorer

Ask HN: To what extent should I be paranoid?

1 points by ge96 ↗ HN
Here's an example project that I had in mind:

- take a screenshot of my screens at some interval

- use OCR (library) to figure out some context of what I was doing

My tin-foil hat comes on "can I trust that library" that I used to do OCR. I would use something like python Tesseract.

Which idk watch your network I guess if it phones home.

This would all run locally but it's that thing of convenience/using something off the shelf that works. More hardcore "safer" would be use OpenCV on its own and do my own image interpretation which would mean longer time to achieve the original goal.

Of course easiest solution is just "write something down" everyday.

I just was wondering what did I do the last week/how was I productive.

2 comments

[ 7.6 ms ] story [ 19.1 ms ] thread
raxxorrax
OCR from scratch is pretty difficult, believe me, I have tried and even with OpenCV support you would need months or years (don't know if openCV is shipped with a OCR solution these days). Even if you restrict yourself to a small subset of characters and fonts you need to solve a lot of problems. There are really good ready to use solutions though. Tried the cloud service from ABBYY which nets really good results, but I guess you wouldn't want that.

The project is an interesting idea though. Would save time in writing all those protocols...

To answer the question, I think being paranoid about such things is very healthy.

ge96
Yeah I don't have a vision processing background so it would be brutal if I tried from scratch. It's just that idea that someone could be peaking into your "soul" ha, if they can screenshot your screen/which can display whatever.

It would definitely need to be a local program.

Edit: another thing is I want to use a Tensor Flow JS model and want to make sure whatever is passed into it isn't also getting sent out somewhere.