1

I'm trying to use the tensorflow object_detection API to analyze individual pathways within a building. The recordings I have access to are really sensitive data and must stay absolutely private.

Is there any chance that the security of this information could be compromised if I read the videos using Python in a Jupyter Notebook? Specifically, I'm worried that the data will be sent to a remote server somewhere where (potentially) other people can view it. Is there a way to make sure that the data remains local?

divibisan
  • 11,659
  • 11
  • 40
  • 58
Janjouneh
  • 145
  • 1
  • 2
  • 9
  • No, but then again you can [never be sure](http://wiki.c2.com/?TheKenThompsonHack) – Dominic Price Jun 19 '18 at 18:27
  • Do you know where I could have this info ? – Janjouneh Jun 19 '18 at 19:09
  • 1
    Can you clarify what sort of things you are worried about? For example, are you worried that the data you use in your program is going to be silently sent to a remote server somewhere where other people can view it? If so, then neither python or any other reputable programming language does this, any data you use in a program is held locally in your computer’s memory when you are running the program and freed afterwards. – Dominic Price Jun 19 '18 at 19:30
  • Thank you, that's exactly what I was concerned about. My intership tutor wants me to give her proof that nobody except us will have access to the data I'll analyze. I just don't know where I can find that proof... And are you sure the data are held locally ? Bc when I launch the Jupyter Notebook it opens up within a internet explorer page – Janjouneh Jun 19 '18 at 19:54
  • @Dominic Price, "freed" does not mean "erased". Memory (and data) can still be read long after the program has terminated. – OregonJim Jun 19 '18 at 23:14
  • @Janjourneh, you cannot provide "proof", because no data is ever 100% secure. The best you can do, from a practical standpoint, is host the data on a computer that is not connected to ANY network, has no removable drives or USB ports, and is located in a secured area. – OregonJim Jun 19 '18 at 23:24
  • @OregonJim, this practical solution doesn't seem very doable though... Let's say, I have the data on an external hard drive. What are the best precautions I could take to avoid leaks and what risks will I have to take ? I really have to know what I'm doing here because I am legally responsible for what happen – Janjouneh Jun 20 '18 at 07:35
  • @OregonJim Your best defence is to make sure you have good antivirus software and run it regularly during the course of your project. You can also install a VM (there are plenty of walkthroughs for how to do this) which isn't connected to the internet and run Python from within that; in order to steal your data someone would have to be constantly uploading the entire contents of the VM's RAM which you could easily check for. Point out to your tutor that Python is regularly used by security professionals. – Dominic Price Jun 20 '18 at 14:09
  • Thank you so much ! I was starting to think I was not going to be able to do what I wanted! – Janjouneh Jun 20 '18 at 14:48
  • I'm still concerned though bc I have the following lines in my code : MODEL_NAME = 'ssd_mobilenet_v1_coco_11_06_2017' MODEL_FILE = MODEL_NAME + '.tar.gz' DOWNLOAD_BASE = 'http://download.tensorflow.org/models/object_detection/' Can I download the models on internet and then run the program within a VM ? – Janjouneh Jun 20 '18 at 15:56
  • Do you have any VM to recommend me ? – Janjouneh Jun 21 '18 at 09:58

0 Answers0