I reverse-engineered Google DeepMind's SynthID — the invisible watermark
baked into every Gemini image.
No access to their encoder. No proprietary code. Just signal processing,
FFT analysis, and stubbornness.
90% detection accuracy. 91% phase coherence drop on removal. 43+ dB PSNR.
Built this alone. For fun. While unemployed.
→ https://t.co/hilkBtTzNf
If this helped you or your team, I'd genuinely appreciate a coffee:
☕ https://t.co/yBQO9mKbJw
Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: https://t.co/CDSQ8HpZoc
Channel 4 Encro: But there was no mention that the Encro evidence itself was highly problematic. French refused to allow the decryption method to tested, the evidence itself showed deep faults and may not have been admissible. https://t.co/vSL2SdpI6y
There's an entire parallel scientific corpus most western researches never see.
Today i'm launching https://t.co/DSD2mZtuap, a fully automated translation pipeline of all Chinese preprints, including the figures, to make that available.
Studying tip:
1. Go to https://t.co/fOYUV0kbbJ
2. Study the bug in detail
3. Study the attack vector
4. Study the protocol where the bug is found
5. Repeat every day and you will progress really fast
Imagine receiving a normal WhatsApp message from someone… and later discovering that the message secretly contained their exact location, even though they never shared it.
That’s exactly what happened during a recent forensic extraction I performed on my iPhone 12 Pro Max.
During the analysis, I found a I decided to pick a message I received from @RedHatPentester, on 3rd September 2025 at 7:11 AM.
Nothing unusual at first glance just a regular text.
But deep inside the message metadata, the phone had silently logged:
@RedHatPentester exact location at the moment he sent that message.
He didn’t share it intentionally.
I didn’t request it.
Yet the device recorded it automatically.
This was extracted directly from my own phone, meaning:
If your location is turned ON while chatting on WhatsApp, your exact location can be extracted from someone else’s device if theirs undergoes forensic imaging.
Most people have absolutely no idea this happens.
But this was only the beginning.
Also, every single file created on the device, ie: photos, videos, screenshots, recordings had the exact location of where I was when that file was created.
The phone automatically logged precise GPS coordinates for each media file.
This means investigators can determine where you were at the exact moment you took a picture, recorded a video, captured a screenshot, or created any media on the device.
This level of metadata helps reconstruct movements, timelines, and behaviors with incredible accuracy.
The extraction revealed far more than hidden location data and remember, this phone was NOT jailbroken.
Here’s what else was recovered:
1. Full Synchronized Accounts & Passwords
The extraction pulled:
•URLs
•Usernames
•Passwords
•Stored login metadata
Basically, every synchronized password ever used on the device all recovered without jailbreak.
2. Complete Application Logs & Histories
Every installed application had:
✔ Detailed logs
✔ Usage history
✔ Internal data
✔ Metadata
Even apps considered “secure” or “encrypted” still left behind recoverable traces.
3. Full WhatsApp Data, Including Group Histories
WhatsApp revealed more than most users realize:
•Full history of every group ever joined
•Date each group was created
•Who created it
•Date I was added
•Group metadata even after you exit the group
This is critical in investigations because a suspect cannot deny belonging to a group when the device itself retains:
📌 Creation date
📌 Creator identity
📌 Join date
📌 Participation timeline
Even if they left the group years ago.
4. Message-Level Location Metadata
The iPhone logged exact sender locations at the moment messages were typed and sent just like what happened with @RedHatPentester message.
Most people never see this.
Investigators do.
Why This Matters
Every phone tells a story.
Every app keeps footprints.
Every message carries more than text.
This extraction proves that even without a jailbreak, investigators can discover:
✔ Locations
✔ Passwords
✔ Group associations
✔ Message histories
✔ Detailed app activity
✔ Metadata most users never realize exists
Digital devices rarely forget even when the user does.
The Call for Papers is now open for the FOSDEM 2026 Open Source Digital Forensics devroom—submit your proposal today. #digitalforensics https://t.co/knJmbuPXqW
“It is clear there is an intent to destroy the Palestinians in Gaza through acts that meet the criteria set forth in the Genocide Convention.”
UN Independent International Commission of Inquiry's report shows #Israel has committed #genocide in #Gaza.
https://t.co/UGjepPwmwo
دمّر الاحتلال معهدهم وقتل طلابه وشرّدهم..
فرقة "طيور غزة تُغنّي" تكافح للاستمرار بآلات بسيطة وأخرى مصنوعة بما يتيسر بين الخيام وتحت قصف الاحتلال المستمر وطنين مسيّراته الذي لا يتوقف.
Had a mentor tell me once: "When you're feeling overwhelmed, there are only two things you should do: get organized and get to work. The rest is just noise. Peace is found in progress."
Developers consistently fail to realise that understanding the problem is 90% of the effort, and writing code is only 10%
They compound that mistake when using generative AI, making the writing code easier at the expense of understanding
Use multiple AIs to explore the problem
My fav little-known trick is to test various syscalls fail with strace fault injection, like:
$ strace -e trace=clone -e fault=clone:error=EAGAIN
https://t.co/au0aM3m13c