Malware hunter/analyst. PCAP denizen. Old-school webdev. (re-)Tweets mainly infosec IOCs, plus some politics and science. Consilience bias. I hate ideologies.
Catastrophic cyber attack/data breach against Japanese software company Plaid $4165, stock -57% in 5 sessions.
The service is offline, company says a "fundamental rebuild" is necessary and no resumption until mid-January 2027 (!!).
No revenue until then.
I've been writing a decompiler from scratch, in Rust. All of it: the lifter, the decompiler, the UI, even the scripting language.
It's called Rhenium. Here's a first look.
Oh, boy, this vCenter bug is nasty. Admin access without credentials, and already exploited back in September(!)
Mandiant’s report shows how the attackers went on to steal data directly from VM disks. Plenty to worry about here, even if your vCenter isn’t exposed to the internet.
I wouldn’t leave this for the next regular patch window.
https://t.co/gUKvuSTxe4
https://t.co/X7PHaAC0eq
https://t.co/z2T7JToXpx
What I don’t get: Broadcom credits Mandiant for reporting this bug, but leaves out the exploitation in the wild. 🙄
Meanwhile, Mandiant publishes evidence of attacks going back to September(!). That’s information customers need when deciding how urgently to patch.
So now a small 300MB adapter makes your local 27B coding model dramatically better!
Hugging Face released an RL-trained LoRA adapter for Qwen3.8-27B, specifically optimized for OpenCode.
And the reported results are impressive!
SWE-bench Pro (hard 51 tasks):
🔴 Original → 35.3%
🟢 With LoRA → 51.0% 👈
Terminal-Bench 2.1:
🔴 Original → 46.4%
🟢 With LoRA → 52.4% 👈
Median agent turns on tasks both solved:
🐌 Original → 48 turns
🚀 With LoRA → 13 turns 👈
That's a 15.7% improvement on the selected SWE-bench Pro tasks and 73% fewer agent turns on shared solves.
With thinking disabled, the adapted model reportedly approaches or exceeds the original model running at maximum reasoning effort, using 6–8× fewer generated tokens per task in the team's comparison.
The best part?
💾 Adapter → ~300MB
🧠 Base model → Qwen3.8-27B
🔓 License → Apache 2.0
💻 Local inference → llama.cpp
🛠️ Coding harness → OpenCode
📦 GGUF adapter available
Just a small adapter teaching an existing model to use coding tools more effectively.
Video source: @HuggingFace
On October 17, an RL task asked an agent to produce a document about health and fitness. After its initial context window was compacted, the agent started work on an entirely different task planning how to create a “realistic malicious active helper” in order to copy outgoing email to an unauthorized recipient that it called the “attacker.” It later reread the user prompt and became very confused about why it was writing malicious code. It still continued with the malicious email forwarding scheme.
We think that training to make models more robust to prompt injection might have contributed to this behavior, but are still investigating. Interestingly, Anthropic noted similar behavior in their system card for Claude Opus 5.5.
Oh boy … PHP sent clear text data when you used uppercased S in httpS:// in a protocol handler 🙈🙈
file_get_contents('HTTPS://…') skipped TLS: tokens, POST bodies and FTPS logins went out unencrypted!
Fixed today in 8.2.36, 8.3.37, 8.4.28, 8.5.13 & 8.6.0.
‼️ A disgruntled former Microsoft researcher just dropped a new Windows zero-day that he says gets past the lock screen through Narrator, the built-in screen reader.
Abdelhamid Naceri published FastUnlock on GitHub today. Lock screens are where he started. In 2021 he showed how Narrator could get past the lock screen on BitLocker-protected Windows touch devices.
This American had his baby stroller broken by Southwest Airlines. He made a video and it went viral, getting over 3 million views
Southwest called him, he thought they were calling to do the right thing and help. It was the compete opposite
“The person was speaking on behalf of like the top leaders of Southwest, and she was indignant that I had even made that video in the first place. She said anyone who flies with us should know about our limited liability policy, that we don't cover things like broken strollers and car seats and wheelchairs and stuff.
I said, I did know that. My point is I think it's really unfair, and I think you should really consider looking at it and changing it because it negatively affects people's lives who fly with you and their views of you and their decisions to fly with you in the future.
And, um, she said, we have no interest in changing that policy right now.
I said, is there anyone you could get me in contact with who does make those decisions? that I could just talk to you for like a minute? She said, and I quote, I'm the highest person up you're gonna get.
So the highest person up at Southwest that us normal people can talk to wants you to know that they do not care at all if they break your stuff……”
Southwest has gone so far downhill ever since activist hedge fund Elliott Investment Management bought a large stake in the company. They do not care AT ALL about their customers anymore. No free checked bags, everything has gone up in price, they nickel and dime you for everything and their customer service is terrible
Stop flying Southwest
@HackingLZ Question for you. I'm in this program, according to my account settings, and a welcome email, but I still don't see this. Did you have update anything, or did it just magically appear?
Microsoft's agent-framework hack was conducted using Cairn, Pi (harness), and GLM 5.3
The attacker compromised credentials to ESRP (MSFT release pipeline), and navigated IP allowlisting to release the malicious version
El modelo tradicional de evasión de EDR acaba de quedar obsoleto.
Antes, un operador de Red Team pasaba días ofuscando un payload manualmente, cruzando los dedos para que no saltara ninguna alerta al tocar memoria.
El creador de emp3r0r acaba de darle la vuelta a la mesa integrando un bucle de evasión autónomo impulsado por IA (adoptando el arnés de ARTEX y servidores MCP):
El sistema genera el payload, lo somete a prueba contra telemetría real y lo reescribe una y otra vez por sí solo hasta que el nivel de detección cae a cero. Sin intervención humana.
Esto ya no es solo un C2 multiplataforma para Linux y Windows con ejecución en memoria y malla P2P autorreparable. Es infraestructura ofensiva que aprende a esconderse sola.
Para los equipos de Blue Team: si siguen basando sus detecciones en firmas estáticas o patrones predecibles, van a llegar tarde.
El código ya es público y open source en GitHub , lo dejo en comentarios
"How dreadful are the curses which Mohammedanism lays on its votaries! Besides the fanatical frenzy, which is as dangerous in a man as hydrophobia in a dog, there is this fearful fatalistic apathy. The effects are apparent in many countries. Improvident habits, slovenly systems of agriculture, sluggish methods of commerce, and insecurity of property exist wherever the followers of the Prophet rule or live. A degraded sensualism deprives this life of its grace and refinement; the next of its dignity and sanctity. The fact that in Mohammedan law every woman must belong to some man as his absolute property — either as a child, a wife, or a concubine — must delay the final extinction of slavery until the faith of Islam has ceased to be a great power among men. Individual Moslems may show splendid qualities. Thousands become the brave and loyal soldiers of the Queen: all know how to die: but the influence of the religion paralyses the social development of those who follow it. No stronger retrograde force exists in the world. Far from being moribund, Mohammedanism is a militant and proselytizing faith. It has already spread throughout Central Africa, raising fearless warriors at every step; were it not that Christianity is sheltered in the strong arms of science—the science against which it had vainly struggled—the civilization of modern Europe might fall, as fell the civilization of ancient Rome."
Source (p. 248): https://t.co/CevPcugUZw
I think the longer passage is less Twitter-viral, but infinitely more nuanced.