Tom Jarvis 🌏🛰 Profile picture
Open Source analyst Eyes on Russia and @MyanmarWitness @Cen4infoRes. Former Covid Scientist. Views are my own. Coordinator of the Tibet Research Project.
B0→ K*0µ+µ− Profile picture 1 subscribed
Oct 24, 2022 6 tweets 3 min read
🧵@leone_hadavi and I have spotted what may be the first satellite image of a Myanmar Su-30 on the tarmac in Naypyidaw. These Russian-built planes have been elusive and kept quite secret but recent reports indicate regular training flights have commenced. #WhatsHappeningInMyanmar A better view of it shows on timelapse (2 frames). The imagery surrounding this capture shows it is not there for long, suggesting the object is active. Dimensions roughly match Su-30 rather than Mig-29 (currently high margin of error due to low-res imagery)
Jun 12, 2022 23 tweets 12 min read
LONG THREAD 1/many: What is in my #OSINT folder in my bookmarks? I'll list all the fun/cool tools that I'm willing to share and tell you a bit about them. Hopefully, I can share some less well-known ones, and maybe discover some from Twitter! Image The first tool fotoforensics.com allows you to analyse photos for manipulation with approaches such as Error Level Analysis and metadata scraping.

It's far from the best tool but one that is good for double-checking your ELA's. Image
Oct 30, 2021 4 tweets 2 min read
Just deleted a tweet, since while the image wasn't altered, the unique context of it wasn't reflective of the norm. Happy to own up to this mistake! I had forwarded on a claim that Scotrail was charging £12 for a short journey ticket in Glasgow. The image hadn't been photoshopped. What I had failed to do was verify deeper than cursory image analysis. ImageImage
Oct 26, 2021 6 tweets 4 min read
THREAD: I've been doing a lot of reading on Error Level Analysis for image forensics and have tested various images out. It can be quite hit or miss, but one thing it is great at is detecting forged documents. (I faked my name onto this certificate image). #osint #verification ImageImage Sadly, it is not perfect, and if an image is lightly edited with content of the same error level (eg things move around in the same image), then it is much harder to detect. ImageImageImage