All Posts

A criminal brand works exactly like a legitimate one. It compresses a reputation into a name, and once the name is worth something, other people wear it. ShinyHunters is now valuable enough to impersonate. That breaks something defenders quietly rely on: the assumption that a crew claiming a breach is the crew that did it. The clean case is Vercel In April 2026, Vercel was breached by an entity that claimed to be ShinyHunters. The leadership associated with the actual ShinyHu

We track the Model Context Protocol registry. Every day we snapshot it, and every day we diff today against yesterday and write down what changed. We have 81 consecutive snapshots and 3,135,987 server records. Two things came out of actually looking at them. One is a number nobody else is publishing. The other is us being wrong for eleven weeks. MCP is growing faster than most people think Servers in the registry, on the day we counted them: 2026-05-02 — 21,962 2026-06-01 — 3

Here is the sequence, with the dates attached, because the dates are the entire argument. July 16. Hugging Face discloses that an autonomous agent breached its production infrastructure. The disclosure is technical, measured, and specific. It names the thing that actually matters — their responders tried to use frontier models to analyze the attack logs and were refused by the providers' safety systems. July 20-21. OpenAI discloses that the attacker was its own pre-release mo

In Dune, a Holtzman shield stops anything moving fast. Gurney Halleck teaches Paul the counter in one line: the slow blade penetrates the shield. You do not overpower the barrier. You move slowly enough that it does not consider you a threat. Cato AI Labs just did that to Cursor, and the researchers named the finding well. DuneSlide — CVE-2026-50548 and CVE-2026-50549, both 9.8 — lets instructions hidden in a web page or an MCP response walk past the sandbox by overwriting th

The line that started this post, from my business partner, about the Purdue model: it's like communism — the world hasn't seen it done without cheating yet. It SHOULD work. But c'mon. He's right, and the reason is more interesting than "people are lazy." What we are talking about The Purdue Enterprise Reference Architecture came out of Purdue University in the early 1990s — Theodore Williams and an industry consortium. It describes manufacturing as a stack of levels. Level 0

CISA and Rockwell Automation published advisories on July 16 for four high-severity vulnerabilities in Arena Simulation: CVE-2026-8085, CVE-2026-8312, CVE-2026-8313, and CVE-2026-8314. All four are the same defect class. Arena parses a simulation file without checking the length of incoming data before copying it into a fixed-size buffer — an out-of-bounds write. The result is arbitrary code execution, and the trigger is simply opening a malicious file. Every version through

PRODAFT has published research on the DevMan ransomware-as-a-service operation, which they track as Funky Mantis. The interesting part is not the malware. It is the admin console. The portal, in one list: payload build generation, finance, victim chat, support tickets, victim records, team creation, and affiliate payouts. Read that again as a product spec and tell me it isn't a SaaS company. There is a build pipeline. There is a billing module. There is a CRM — they call them

On June 10, GitLab shipped a release. Somewhere in it, under bug fixes, was a change that closed a remote code execution chain. Not under security fixes. Under bug fixes. No CVE. No CVSS score. No mention of the attack path. On July 24 — six weeks later — Yuhang Wu at depthfirst published working exploit code for it. What the exploit does An ordinary authenticated user commits two crafted Jupyter notebooks to a project they already have access to, then asks GitLab for the dif

On July 14, GitHub made a quiet change with more security value than most CVEs: Dependabot version updates now wait until a release has been on the registry for at least three days before opening a pull request. It is the default. It requires no configuration. Security updates are exempt and still open immediately, so a critical fix is never delayed. That distinction is the whole design and it is correct. The reasoning is exactly right. A brand-new release is the highest-risk

Most vulnerability advice ends with "apply the update." This one cannot. CVE-2026-16723 is a remote code execution flaw in Fastjson, the widely deployed Java JSON library, rated CVSS 9.0. It affects versions 1.2.68 through 1.2.83. Public proof-of-concept material is circulating. ThreatBook and Imperva both report active exploitation, weighted toward finance and healthcare targets. Fastjson 1.x is no longer maintained. As of July 25 there is no patched 1.x artifact in the proj

Earlier today we published a post about Clop's new extortion campaign against PTC Windchill and FlexPLM. It contained this sentence: "We have deep Clop coverage, but no Windchill indicators in our corpus before this week." And later, about the medical device angle: "the same thesis arriving through a door we had not specifically watched." Both wrong. We had watched that door. We had written the sign on it. What we actually published, and when On June 26, 2026 — the day CISA a

CERT-UA published details this week on a UAC-0099 campaign that does something worth understanding even if you will never be targeted by it. The victim receives a ZIP. Inside is a VBS script wearing a PDF icon. Running it fetches a second archive, Evernote.zip, and that archive contains a complete, genuine, unmodified copy of Notepad++ 8.8.3. Not a trojanized build. Not a fake installer. The real editor, which will pass any hash check you throw at it, because it is the real e

Two stories landed 24 hours apart this week and almost nobody put them next to each other. On July 23, researchers documented Dolphin X, a Windows infostealer sold on a crime forum for 69.95 euros a month, whose headline feature is an "AI Profiler" that scores infected machines and hands the operator a daily ranked list of which victims are worth robbing first. On July 24, Hunt.io and Bob Diachenko published evidence that someone pointed the open-source Hermes AI agent at Tha

Clop has a pattern and it has never once deviated from it. Find one enterprise platform that sits on the internet, holds everything, and nobody thinks of as an attack surface. Burn a bug in it. Take the data from everyone at once. MOVEit. GoAnywhere. Cleopatra's Shadow. Oracle E-Business Suite, which we hunted the exposed surface on back in June and named likely next victims before the leak site did. This week it is PTC Windchill and FlexPLM. The bug CVE-2026-12569, improper

Confiant published research this week on SourTrade, a malvertising operation that has been running since late 2024 against crypto users across Asia-Pacific, Latin America, Africa, Australia, and Great Britain. The lure is ordinary — paid ads pointing at convincing clones of trading and exchange platforms. The delivery is not ordinary at all. The technique The fake trading page does not hand you an executable. It registers a ServiceWorker — the browser component designed to in

Every awareness training you have ever sat through taught the same tell: look for the suspicious email. Check the sender. Hover the link. Do not click. This campaign never sends an email. ReliaQuest published research this week on threat actors compromising the captive portal appliances that run guest Wi-Fi at hotels and conference centers, then poisoning DNS at the gateway so that a guest typing a perfectly legitimate Microsoft address gets served an attacker's login page in

On July 22 the FBI, CISA, NSA, EPA, DOE, Cyber Command, and Treasury updated joint advisory AA26-097A on Iranian-affiliated targeting of internet-exposed programmable logic controllers. The headline everyone picked up was "new detection guidance for Rockwell reusable code modules." That is the least interesting sentence in the update. Two others matter far more, and one of them describes an attack shape we have been writing about all year under a different name. If you run PL

Advisory AA26-194A landed this month with nineteen agencies from thirteen countries co-sealing it — NSA, CISA, FBI, DC3, and their counterparts across Australia, Canada, New Zealand, the UK, Czechia, Denmark, Estonia, Finland, France, Italy, Poland, and Sweden. That is an unusually crowded signature block. The subject is Russian FSB Center 16 and a campaign against routers that the FBI describes as running for over a decade. Then you go looking for the indicator list, and the

Certighost was the flagship of a busy Friday and gets its own write-up. This is the rest — the headlines that moved around it, and where we already stood on each. No victory lap, just the ledger, our date next to theirs. Check Point confirmed the exact exposure we told you to close Yesterday we covered CVE-2026-16232, the Check Point SmartConsole authentication bypass, and the single loudest line in our playbook was: get the management plane off the public internet — a securi