What Problem Solving Actually Looks Like
The proof of concept images for the vending machine hack were lost. How did I get them back?
August 9, 2020: I was starting to piece together my last week’s post, when I opened Google Photos for the images to be included as proof-of-concept. Unfortunately, I couldn’t find them anymore. Then I remembered factory resetting my phone two weeks after, so they must have been wiped since I do not remember backing them up. But you saw my last post, and it has those images. How did I get them back?!… Through an HTTP Archive file and some amazing Linux tools.
I posted few photos of the hack on Instagram stories some 4000 years ago. Luckily, I added them to the profile highlights as well.
Now that they’re stored on Instagram servers, I just need to find a way to get them on my computer since there is no “Save to Device” option in the app menus. I quickly logged in to Instagram web to see if they have the highlights function there and if so, I can get links to those images using the most amazing piece of hacking software there is… the Developers Tools.
I opened the console to see GET requests with the source image request URLs. Awesome! I can download the images now,… but should I? They were NINE images, NINE links to be clicked and saved manually using more clicks. The math is too much! I decided to look for a more automated solution in case there’s ever a need to scale up. Enter HAR.
This HTTP Archive file, which can be exported from the Networks tab, is a JSON file containing detailed performance data of the browser about the webpage it loads. This would include the image requests URLs I’m interested in. I’ve been looking for a way to collect request URLs like this for ever, and I’ve finally found it during this engagement.
After analyzing the JSON structure, I found which request URL values I need. It’s the one with host as ‘instagram.fdel11–1.fna.fbcdn.net’. I used jq for parsing through this JSON data, all 15,566 lines of it. I thought about using grep or awk or sed but I suppose they would have required a lot more tweaking. Plus I wanted to get better at jq, since I’ve only used it once before.
I spent some time reading up on how to traverse arrays, creating a regex for string matching the URL, also trying to resolve a recurring “error: test is not defined” . Let me know if you managed to solve this because I couldn’t get it working on my terminal. So I resorted to using their online playground for testing… A one-line bash command for extracting direct image URLs from the HAR file. And another one to download them.
At this point I had everything I needed, complete with a bash script to go over the URLs, and download the images using wget. Because everything looks cool in script?! Take a look for your experimental pleasures :p
This is an example of how problem solving works in the real world. Taking existing random knowledge bits, sometimes learning new bits as well and connecting them in a clever way to get results. Click below to see my last week’s post where I used these images for proof of concept.
