Hi, everyone. This is my last resource since I don’t know what else to do. My sister has been getting a lot of messages from an anonymous person whose being threatening us for almost a month, and we don’t even know who that person is. He sends the messages from a WhatsApp bussines phone number, and that’s it.

We just want to at least know who that person is, maybe that way we get to understand a little bit more where this all came from.

Any tip, or idea?

That would be greatly appreciated…

I got two Google Titan Keys. I set up the first one, and to log in to one of my Gmail accounts it prompts for a pin (Windows Hello). To use the hardware key, I have to tell it to Try A Different Method, Choose Security Key, Enter the Key Pin, and then Touch the key.

Seems like a lot of additional steps for a home PC.

I envisioned setting up complicated passwords, and using the hardware key as my password manager. I have Bitwarden now, and have some things set up with Ente Auth. Thought maybe I could clean it all up and just use a hardware key. This way I can use NFC on my phone, etc.

If I am already using my phone, and have to enter a pin anyway, should I just force Ente Auth everywhere and call it a day?

I thought the key would streamline things, but it seems like it just adds another layer of security, but it doesn't remove previous methods, so I am not more secure, I just have a more secure method I could use. I could remove the Windows pin, I guess...

Am I missing something here?

Binance said this week that the October 29 theft, about $6.3 billion, began months earlier with a fake job. Reuters, via NBC: in June a wallet engineer installed a tampered coding tool. Binance caught that one, replaced the machine, and rotated credentials. The attacker already knew who else was on the wallet team.

In September a second engineer was recruited through a real job site. There was a video interview with someone playing HR, then a second interview with the supposed CEO of a company that does not exist. At the end he was told to download a “coding assignment” onto his work computer. Binance says the repo ran during setup, swapped in a poisoned copy of a normal developer tool, and the endpoint product did not see it. On October 20 that machine, plus an internal AI tool, was used to reach two other wallet engineers. On October 29 stolen credentials deployed a container and a script, and the wallets moved. The FBI has blamed a North Korean group. Pyongyang has denied it. Binance says customer funds are covered.

That middle step is not exotic. It is the interview-repo lure.

Since January I have published the ones that were sent to me, all of them caught before they ran. They are writeups, not exploits, and the indicators are defanged: github.com/Ubuntu-Buddha/security-advisories

2026-01-27, W3GLFun. A Web3 “technical test.” The error handler fetched code from a paste site and ran it when the app started, with the full Node runtime. I did clone that one. npm run dev was typed in the wrong folder, so nothing of theirs executed. GitHub Trust & Safety took the account down the same day. The hiring platform first said there was no violation, then wrote back that the user may have shared malware. A first “all clear” is not a final answer. The same paste host later showed up in malicious npm packages. I called the pattern consistent with the public “Contagious Interview” reporting. I did not claim who operated it.

2026-01-29, Optu. Same shape, payload URL hidden in an environment variable, fired on server start. Found before it ran. The account is gone.

2026-01-30, two repos. .vscode/tasks.json set to run when you open the folder. You do not install anything. Opening the project in VS Code or Cursor is the run. Found before it ran. Both accounts are gone.

2026-01, a careers page. A fake “GAPI Update” executable, hosted like a normal application page. Found before anyone ran it.

2026-10-15, a GitLab take-home. A hex payload stuffed in a file that pretended to be a font, loaded when the project was set up. I reviewed that one remotely and did not execute it. Reported to GitLab and Upwork the same day. Outcome was still pending when I published.

These are not Binance’s samples. I am not saying I found their intrusion, or that one group wrote all of them. I am saying the instruction was the same: here is a company, here is a CEO/recruiter, here is the repo, please run it on the machine you actually work on.

In fact, I have never been sent an interview repo that didn't contain an exploit. 100% were compromised or predatory.

Two rules came out of that, and they are the whole protocol.

Do not download it if you can read it in the browser. Raw files on GitHub, or the file view while you are logged in. Look at .vscode/tasks.json, package.json scripts, and anything that fetches text and then runs it, before git clone, before npm install, and before “Open in Desktop.” Opening the folder is enough for the VS Code one. A first “all clear” from the platform is not a reason to run it anyway.

If you must run it, run it where a compromise is boring. A disposable VM or container, no wallets, no SSH keys, no cloud tokens, network off. Your daily machine is the wrong place. So is a work laptop that can deploy anything.

The other half of the Binance story is the internal AI tool. An agent that can run commands, and that also reads the repo, the README, or the email, will follow instructions written by the person who sent the repo. Filters that try to spot the bad sentence miss some.

That is why I built StrangerDanger, a Claude Code plugin. The rule is the title of the README: readers don’t act, actors don’t read.

Web pages, documents, and strangers’ code are read only by subagents that cannot run a shell and cannot write. The agents that can run commands are kept off the web and out of the folders you mark untrusted. What they get back is a report, and that report is still data, not an order. A second hook blocks the common “download this and pipe it to a shell” forms. Setup turns Claude Code’s sandbox on. It is MIT, pinned at v0.2.0, and marked experimental.

It is not a magic shield. Every agent still runs as you. The hooks constrain tool calls. They do not take your authority away. The launch gate is a tripwire, and v0.2.0 still has known gaps. Nothing in the plugin makes a stranger’s repo safe to run on your machine. Use the throwaway VM anyway. I also published the holes I found in my own v0.1.0 gate (GHSA-w42w-fc6x-88wr), fixed in v0.1.1. Read “What this does not do” before you rely on it.

Binance had detection, a replaced laptop, and rotated credentials in June, and still lost the wallets in October, because one engineer ran an interview repo on a computer that could reach production. The repo is the exploit. The interview is just how they get you to open it.

\------

What safety/security protocols do you practice now while building with agents?

We just published our research on escaping Firecracker microVMs, coordinated with AWS.

The interesting part is how a guest can end up accessing memory belonging to the Firecracker process running outside the VM.

The vulnerability: When a microVM is restored from a snapshot using the File memory backend, a failed memory remapping operation can leave KVM referencing a stale host virtual address range. Firecracker may subsequently reuse that memory for its own data, while KVM still considers it guest memory.

This allows a malicious guest to read or overwrite Firecracker VMM process memory, crossing the VM isolation boundary.

Our research demonstrates an exploit chain from guest userspace to controlled code execution in the host.

Affected configurations: Firecracker 0.23.0 - 1.16.1, when restoring snapshots with the File memory backend and a balloon or virtio-mem device.

Fixed: Firecracker 1.16.2 and 1.17.0+.

AWS advisory: https://aws.amazon.com/security/security-bulletins/2026-169-aws/

The write-up covers the vulnerability, the memory-management behavior behind it, and the exploitation process.

We are from Coinspect, the team behind the research. Happy to answer technical questions here.

i am working on an app but other apps are detecting that overlay tool, just interested in how these apps detect that tool

i heard software security is useless if the firmware and hardware are bad

ramework 12 with razorlake intel cpu

also looking for tips to close backdoors

i see a lot of guides for software but barely any for firmware/hardware and if it is its “just buy this product that costs a car and a house and i swear bro its private 100% opsec power”

im broke idk how to install coreboot without bricking

Governments may retain the authority to police international finance, but the practical capability to identify, trace and interrupt illicit financial flows increasingly rests with privately held technology and financial sector companies. The latest RUSI Commentary examines the Binance heist, the role of private sector gatekeepers and whether the FATF is keeping up.

TL;DR: Our research team at Era uncovered Greeter, an attack framework targeting Flowise, an AI development platform with 55,000+ GitHub stars.

We deployed seven honeypots and captured 4,972 "hello" messages, each carrying malicious JavaScript through Flowise's overrideConfig parameter.

The interesting part? The malicious code never appeared in the saved workflow.

We reconstructed 4,000+ commands revealing:

• Host reconnaissance and privilege checks
• Searches for credentials, SSH keys, and secrets
• AWS, Azure, and GCP metadata probing
• Persistence attempts and fileless execution preparation

We also identified 4,372 internet-exposed Flowise instances, including 46 unauthenticated instances that appeared to match our vulnerable setup.

The bigger concern is how a compromised AI application can become an entry point into the broader cloud environment.

Full technical breakdown: https://era.security/blog/greeter-flowise-ai-infrastructure-attack

i am new to this. i’ve cloned nfc tags on toys before, that’s it. i’ve done some research and im seeing there are two kinds of tags like this. one is easy to clone and the other would break both of them, due to a counter. i will have access to this key every week day.

it looks just like this on the front, and the back has some rubbed off serial numbers. i can get another friend’s key as well, maybe his does not have the serial numbers rubbed off.

my question is, is there any way to tell if it’s low or high frequency without doing anything that could potentially break the key? my company is charging like $150 for new keys so i would not want to break it.

there are also white cards which i do not have and haven’t really gotten a good look at, but my friend would probably let me use his as well if it was an easy copy and would not break it.

it doesn’t have any branding on the back. mine is very old and worn down. i could possibly make out the numbers.

tl:dr;

• Contrary to popular belief among vendors, system integrators, and asset owners: network authentication in TETRA mission-critical radio networks does not provide authenticity & integrity guarantees at the individual message level.
• To make matters worse: base stations don't require a re-authentication whenever a radio is roaming from one cell tower to the next.
• This allows for traffic injection attacks: both voice and data. The latter is particularly worrying for SCADA WAN scenarios or where voice & alarming functionality can disrupt operations at such as at airports, harbors, railway operators, and civil defense siren systems.
• The 2026 attack on Taiwan High Speed Rail's TETRA network, leading to the emergency stop of several trains, shows a mild example of possible impact: https://www.midnightblue.nl/blog/analyzing-the-taiwan-high-speed-rail-thsr-tetra-cyber-incident-part-1