



Background in OSINT and security,
I’m revisiting an older case involving a group image where faces have been obscured using graphic overlays (likely rasterized and flattened). The image appears to have been recompressed multiple times (e.g., platform upload), and metadata is stripped.
I’m not trying to identify individuals or reverse anonymity, this is strictly about understanding the forensic limits and validating image manipulation.
Current assumption:
Given recompression and rasterized overlays, any underlying facial data is irrecoverable.
What I’m exploring:
Whether compositing can still be reliably detected
via: double JPEG compression artifacts
local noise inconsistencies
boundary detection between original image and overlay regions
Whether PRNU / noise residual analysis is viable at this quality level, or effectively destroyed
What I’ve tried:
ELA-style analysis suggests manipulation but not conclusive
EXIF/metadata, stripped
Reverse image search, no useful matches
Question:
At this point, is there any meaningful forensic approach to validate compositing beyond basic ELA, or is this realistically a dead end due to recompression?
If anyone has experience with forensic tooling (or relevant academic work), I’d appreciate a sanity check on this approach.
I’ve been doing a lot of manual work going through large public image sets (events, protests, archives), and the biggest bottleneck was always the same:
→ scrolling through thousands of photos
→ spotting the same faces again and again
→ re-checking identities manually
So I built a small local tool to speed this up.
What it does:
extracts faces from image folders
clusters similar faces (DBSCAN)
lets you label a cluster once and reuse it
runs fully offline (no APIs, no uploads)
What I found useful:
grouping recurring faces quickly
reducing manual review time
creating candidate sets for further verification
Quick test: ~5000 images → ~15k faces → clustered in a few minutes on my machine
Important:
this is NOT perfect identification
there are false positives (similar faces, lighting, angles)
still requires manual verification
I’m not selling anything right now — just trying to see if this is useful for others doing OSINT or large dataset analysis.
If you’ve dealt with similar problems, I’d love to know:
how you currently handle image-heavy investigations
what breaks in your workflow
If anyone wants to test it on real datasets, I can share access.
I could really use some help. I sold my vehicle to someone who did not register it. Now I'm getting parking tickets sent to me. I have the bill of sale with their name and signature, but the city won't do anything without a home address.
Their name is extremely common so looking them up via traditional methods like the white pages has returned dozens of results around the same age in the same city.
So, is it possible to find their home address with just their name and facebook profile? I have their instagram too. I also know what city they said they live in, but it's a big metropolitan area and they may technically live in an adjacent smaller town.
I suspect this person is probably not malicious. They seemed nice just young and not wealthy. I sold them a beater for cheap and warned them the registration would be expensive because I had it registered non-op. I suspect they got sticker shock at the registration and just never completed it because they couldn’t afford it. I just want to officially inform the city so I don’t tank my credit from unpaid parking tickets.
Appreciate any and all suggestions or help finding this info, thanks!
Mods: if you need proof that the circumstances are as I described I am happy to share pics of the paperwork privately.
WhoCord is used to automate the tedious process of checking which sites registered an email address, finding connected profiles, and generating a security report, It's a Python tool with a web dashboard, supports 700+ websites, and uses only publicly available information.
It can also scan discord urls shared in a server or multiple servers
Everything runs locally, tokens are never stored in plaintext, and it's intended strictly for personal use and authorized testing
GitHub: https://github.com/Siv-nick/WhoCord
Hope it helps others audit their own online presence as much as it helped me
I'm genuinely interested in talking to you. I've come across this work and your stories many times. I just want to hear about you, your journey, etc. I'm curious how you got here, whether you work for the authorities or for yourself, and how legal is it, do they teach this in universities?
I'm not planning to work in this profession to be honest, but I sincerely want to listen to you. If I'm writing in the wrong place, please forgive me, I'm not that experienced in Reddit.
I need an OSINT/SOCMINT expert who can help me find the admin, email, name, phone number or maybe a photo who is running a certain FB page. Please DM for more information.
After years working in private investigation and executive protection, one thing kept coming up.
Most OSINT tools assume you’re sitting at a desk.
Multiple tabs. Time to think. Time to cross-check.
That’s not how a lot of real situations actually happen.
I’ve had moments where I had a name or a plate, and maybe a minute to decide if something didn’t feel right before moving forward.
The information was out there.
But pulling it together fast enough to matter was the problem.
What I kept running into wasn’t lack of data.
It was friction.
Switching tools. Missing connections. Not spotting contradictions until it mattered.
That gap between “available data” and “usable in the moment” always felt bigger than it should be.
For those of you working outside a desk setup, how do you deal with that?
Hi all! I don’t know if this is an ok post to leave here so please lmk if i’m violating any rules! But I was in a hit and run and looking to zoom into a video to see a license plate of the person who did it. My car is totaled and if I don’t find them I get nothing. I know people are going to say “just say enhance at it” or “that’s only in csi” but there are characters there’s just some motion blur and it needs to be upscaled. I have seen some videos where people have done this exact thing with similar photos using video cleaner, amped five etc.
I don’t have a ton of money but please reach out if you are interested. I can give you the material and you can decide if you want to work on it.
Even if you aren’t willing to do it, some layman’s advice on what tools I can use would be great. Thanks so much in advance!!