Lead · Vulnerabilities

The box you bought to hold the perimeter is now the way in, and this time the attacker didn't want your data

CISA put CVE-2026-123039 on the KEV list; the chain runs PHP2Shell for access, Gaia OS for root, a backdoor called TSGA for persistence, and a wiper for the ESXi hosts.

confidence 0.85

CISA added CVE-2026-123039 to the Known Exploited Vulnerabilities catalogue — an incorrect permission assignment for a critical resource affecting Check Point Firewall, Multi-Domain Security Management Server, Scalable Platforms, Security Gateway and Security Management Server, in the Gaia OS DynamicCLI component. The news cycle arrived 7.7 hours later, which is roughly how long it takes a Friday to become a weekend.

The shape of it, from the reporting: PHP2Shell gets initial access to unpatched Security Gateways, CVE-2026-123039 escalates to root, a previously unreported backdoor called TSGA settles in, and then the attackers wipe files on VMware ESXi hosts. Check Point has released emergency hotfixes and says exploitation has been running since late October. A Telegram exploit channel, which is not a source you'd stake a change window on, posted a link to a purported proof of concept and put the start date at 30 October.

Separately, and unproven as a link, an OTX writeup this morning describes adversaries systematically destroying a virtualisation estate over ten days: four ESXi hosts, every virtual machine, both vCenter Servers. That is not a ransom negotiation. That is somebody closing the business down and leaving the lights on, and it is the thing your backup strategy has been quietly promising to handle since 2019.

Why it mattersIf you run Check Point gateways, the hotfix is step one and hunting for TSGA and ESXi file destruction since 30 October is step two; a wiper doesn't give you a negotiation window.

Today’s revelation

A hotfix that nineteen of the twenty-one compromised servers were already running is not a control. It is an alibi with a version number.

— Sturgeon’s Revelations · Probably Fine Daily No. 48

§2

Top stories

The rest of the last 24 hours, with receipts.

Vulnerabilities

Three more KEV entries, and one of them is an IRC daemon, because the internet never actually deletes anything

Alongside Check Point, CISA added CVE-2026-114713 in UnrealIRCd — an improper neutralisation of delimiters that lets any user on a network with two or more linked servers inject server-to-server commands executed with server authority — and then, this morning, CVE-2026-5430 in WSO2 API Control Plane, API Manager, Traffic Manager and Universal Gateway (path traversal to unrestricted file upload to remote code execution) and CVE-2026-71362 in Adobe Commerce and Magento (incorrect authorisation, elevated access to sensitive resources, no user interaction). Meanwhile the ISC reports live exploit attempts against CVE-2026-102895, a trivial authentication bypass in GPUStack affecting all versions before 2.2.0, which was released at the end of June. GPUStack manages clusters of GPUs, which run AI workloads, which is where the money currently lives.

confidence 0.90

Threat actors

N-able pulls N-central offline after 33 flaws, and the hotfix turns out to be a lovely thing to have been running while you were compromised

N-able took the N-central platform offline after the discovery of 33 flaws and shipped a hotfix as recommended mitigation; attacks began shortly after a researcher dropped over 30 zero-days, and N-able patched the unauthenticated ones. Then the arithmetic arrived: attackers infected 21 servers, 19 of them running Hotfix 2, with the new flaw set tracked as 4EverAble2Shell and Agent2Shell, and N-able is now restricting certain N-central features and integrations. In the next lane over, ConnectWise confirmed that abuse of an independent third-party extension let attackers bypass ScreenConnect authentication, with researchers spotting intrusions at nearly 50 organisations; ThreatFox is carrying a ScreenConnect.ClientSetup.msi hosted on a Cloudflare R2 bucket, which is not shown to be the same campaign but is the same idea wearing the same coat.

Why it mattersRemote management tooling is the fastest route from one compromise to all of your customers, and 'we applied the hotfix' is no longer a sentence that closes a ticket.

confidence 0.80

Ransomware

Yesterday's leak-site 'surge' was mostly a posting schedule, and the deadlines are Monday and Tuesday

Pixelleaks put up 14 entries on 13 November against a median of one on other days — a Hong Kong green-building body, a Chinese oncology genomics outfit, Canadian and US financial advisory firms, a welding school — with full-leak timestamps stamped for 2026-11-16 11:45 UTC and 2026-11-17 01:36 UTC. n0n posted 10 at once, including claims against DraftKings, LiveRamp and LEO Pharma that nobody has verified; Play posted nine, Akira six, Qilin five, and The Gentlemen's Atlas escalations all carry one identical 17:50 timestamp. Qilin's five include the University of Texas at Arlington, which has separately warned that personal information was stolen in a cyberattack, so at least one line in the batch has a named human cost behind it. A Darkfeed tally for 2026 puts Qilin on 1171, The Gentleman on 1026 and Akira on 441.

Why it mattersCount victims, not posts: a 14-item day on one leak site is a publishing decision, and the only hard thing in it is the 16-17 November deadline.

confidence 0.65

AI & security

The FBI's complaint about Artex is not that it is clever, it is that it is quick

An FBI flash alert says the open-source hacking tool Artex "significantly reduces the time required to chain attack techniques, with some intrusions progressing from initial targeting to privileged access in under three hours". Dark Reading has a companion case: an attacker whose hand-typed commands failed against a US retailer switched to AI-generated scripts and achieved lateral movement in minutes. On our side of the table, OpenAI unveiled Daybreak Red Team, which pentests web applications with AI agents, and Defense Factory, which investigates findings and prepares patches; Apache's critical mod_http2 race condition CVE-2026-119618 was found by Anthropic and Google. And The Hacker News raises the question nobody's runbook answers: shutting down a rogue agent may not stop it, because queued tasks, active sessions and valid credentials carry on without the model.

Why it mattersIf initial targeting to privileged access can be under three hours, your detection budget is measured in minutes and your approval workflow is measured in days.

confidence 0.80

Malware & IOCs

Someone opened three Trezors, added a circuit board, and sold them anyway

Trezor has confirmed that three Safe 5 wallets sold by the Southeast Asian reseller CryptoBilis contained hidden hardware implants: an extra circuit board that reads the recovery phrase off the device's screen and sends it over the mobile network. A firmware fix is out; it does not undo earlier exposure. The same day, the FBI attributed more than $400 million in cryptocurrency theft over 18 months to a North Korean social engineering operation called FaceStealer, built on fake video calls, deepfakes and stolen webcam clips from over 900 victims, and said it has seized the attack infrastructure — while US authorities separately seized $1.2 billion in cryptocurrency stolen from Binance.

Why it mattersSupply chain assurance for hardware means the chain of custody from factory to hand, not a hologram sticker and a reseller's good word.

confidence 0.85

Policy

Operation Endgame takes 235 servers and 6.4 million stolen logins, and 354 people get a visit

Investigators seized 235 servers and 257 domains behind Gootloader, CastleRAT and StealC, along with about 6.4 million victim credentials, as part of Operation Endgame. In a separate joint international operation, 354 suspects were arrested — vendors and buyers across the Archetyp, Nemesis, Tor2Door, Bohemia and Kingdom dark web marketplaces. German authorities also took down rebuilt SteaIC malware servers for the second time this year, tracing the network to the Virtualine bulletproof hosting provider seized by the FBI in October, with Frankfurt police and the Dutch National High Tech Crime Unit. And the Bundestag passed the biggest overhaul of Germany's spy laws of the postwar era, granting hacking and sabotage powers, by 308 votes to 190 with 71 abstentions.

Why it matters6.4 million credentials recovered means a lot of them are yours, and the takedown does not rotate them for you.

confidence 0.90

Vulnerabilities

PostgreSQL fixes a 22-year-old bug, and the patch is only half the job

PostgreSQL shipped fixes for 52 security flaws, among them CVE-2026-6460 (CVSS 8.8), a type confusion bug 22 years old that lets an authenticated user with object-creation privileges run code on the database server — and full protection also requires rebuilding any C-language extensions installed on the server, which is the part that will be skipped. Elsewhere in the pile: Citrix patched 10 critical NetScaler vulnerabilities, Google fixed 83 flaws in Chrome 157 and Mozilla 99 in Firefox 160, and a critical Red Hat Quay flaw could expose private container images. Microsoft, for its part, confirmed that November's updates are causing Windows activation failures with 0x80041008 errors and may break printing for documents with complex formatting.

Why it matters'Patched' and 'rebuilt the extensions' are different states, and only one of them is actually fixed.

confidence 0.90

Oddities

On Friday the 13th, email went down, air traffic comms rebooted, and the benefits portal chose that evening to be unavailable

The FAA issued a ground stop citing "extraordinary circumstances", with federal officials not saying what caused the communications disruption or whether a cyberattack was being investigated; reporting circulating among responders attributes it to an update to the private communications network connecting air traffic control, operated by L3Harris, which caused a reboot and then a crash. Microsoft, meanwhile, spent two days trying to restore Outlook, Exchange Online and Teams after failed server health checks, with no evidence of a cyberattack disclosed. And somewhere in the middle of all that, one person on infosec.exchange tried to finalise their 2027 benefits at 11pm on the deadline and found Workday down; service was not restored before midnight.

Why it mattersMost of your worst days are change management, not adversaries, and your incident process should be able to tell the difference before the board does.

confidence 0.65
§3

Hidden

The daily deep read: what is under the feeds, not on top of them.

Under the feeds: one real cross-source story, a lot of trackers quoting each other, and a collection gap wearing a disguise.

  1. Exception 01

    Check Point: KEV first, news 7.7 hours later, exploit channel the same evening

    CVE-2026-123039 is the only entity today that appears across CISA KEV, mainstream news and a Telegram exploit channel within a single day, with KEV leading news by 7.7 hours. The KEV and news facts are solid; the PoC link and the 30 October start date come from a low-reliability channel, and the unusual six-digit CVE format is worth a second look before you quote it in a change record.

    confidence 0.60
    Wrong if
    A vendor advisory or NVD record showing a different identifier or no such flaw; the linked PoC proving non-functional or unrelated.
    Watch
    Check Point's advisory, PoC validity, and ESXi wiper IOCs over the next 48 hours.
  2. Exception 02

    The leak-site 'spike' is mostly trackers mirroring trackers

    Pixelleaks (14 entries against a median of one), n0n (10), Play (9) and Akira (6) all burst on 13 November, and the two trackers propagate in a consistent order — ransomware.live typically ahead of ransomlook by 2 to 16 hours. Treat apparent corroboration between them as one source counted twice. The pixelleaks full-leak timestamps of 16 and 17 November are the only hard forward trigger in the set.

    confidence 0.60Speculative
    Wrong if
    Distinct victims with independent first-seen dates confirmed at the group sites; n0n's claims acknowledged by the named organisations.
    Watch
    The 16-17 November pixelleaks deadlines; whether DraftKings, LiveRamp or LEO Pharma respond.
  3. Exception 03

    TSMC: the seller's post came before the confirmation

    A forum actor offered TSMC data for 3 BTC on 8 November; TSMC confirmed on 12 November the exposure of current and former employees' personal data linked to a third-party cloud provider, with no evidence its own systems or customer data were compromised. Claim-before-confirmation is a usable early-warning pattern. The wider Taiwan-semiconductor framing — a Carsem listing by emperador, SilentAir intercepting calls and SMS with Taiwanese subscribers as primary targets — is suggestive, not established.

    confidence 0.55
    Wrong if
    The exposed data proving materially different from what was offered, or the Carsem listing having no Taiwan connection.
    Watch
    Further Taiwan semiconductor supply-chain listings.
  4. Exception 04

    NVD and Atlas both stop at the same minute, which is a pipeline, not a quiet week

    NVD and Atlas show no items after 2026-11-13T00:20, and Vexy Ransomware's listings stop at the same moment. A shared cutoff across three unrelated producers is an ingest gap. Do not read the silence as a drop in vulnerability publication or actor activity.

    confidence 0.50Speculative
    Wrong if
    Items from NVD or Atlas timestamped after 13 November 00:20 turning up in the next pull.
    Watch
    The next ingest cycle; Vexy's next listing.

Nothing to seeNo genuine timing clusters today. The ThreatFox batch all shares one ingest timestamp, so no first-seen timing can be read from it, and the AdaptixC2 and Cobalt Strike infrastructure overlap is plausible but unproven. Telegram leads Mastodon by 13 to 19 hours on the same hashtags, so Mastodon counts are an echo, not a second source. The NVIDIA chatter and the Claude Mythos mention are uncorroborated noise.

Deep read: 468 items read

§4

Forecast

Dated, falsifiable calls. We mark our own homework in public.

Today’s calls

  1. DueWed 18 Nov4 days

    Pixelleaks publishes at least one of the full data sets scheduled for 16-17 November, visible on ransomlook.

    Wrong ifNo pixelleaks publication appears after the stated deadlines, or the entries are removed.

    f-2026-11-14-1

    Confidence0.60
  2. DueSat 28 Nov14 days

    At least one N-central vulnerability from the 4EverAble2Shell or Agent2Shell set is added to the CISA KEV catalogue.

    Wrong ifNo N-able or N-central CVE appears in KEV by that date.

    f-2026-11-14-2

    Confidence0.35
  3. DueSat 21 Nov7 days

    Qilin posts five or more new victims across the ransomware trackers, holding its position at the top of the 2026 tally.

    Wrong ifFewer than five new Qilin listings appear on ransomlook or ransomware.live by that date.

    f-2026-11-14-3

    Confidence0.75

Scorecard

5 right · 1 wrong · 3 open

  1. open

    No public vendor or CISA advisory ties the Ivanti Connect Secure exploitation chain to a specific named threat actor or ransomware group by 2026-11-12.

    Called Mon 2 Nov · due Thu 12 Nov

    No items in today's feed tie the Ivanti CVE-2026-111357 chain to a named threat actor or ransomware group; deadline 2026-11-12 passed with no such attribution appearing.

  2. open

    Ubuntu moves the IPv4 TCP use-after-free kernel fixes out of its opt-in testing channel into regular updates for its three newest LTS releases by 14 November 2026.

    Called Wed 4 Nov · due Sat 14 Nov

    No feed item today confirms or denies Ubuntu moving the IPv4 TCP use-after-free kernel fixes from testing to regular updates by 14 November; deadline is today but nothing settles it.

  3. wrong

    At least one further airline, travel or hospitality company publicly discloses a breach traced to Amadeus IT Group.

    Called Thu 5 Nov · due Sat 14 Nov

    No new Amadeus IT Group-linked airline, travel or hospitality breach disclosure appears in today's feed items by the 14 November deadline.

  4. open

    IHG Hotels & Resorts does not publicly confirm the BYOD intrusion claim.

    Called Thu 5 Nov · due Thu 12 Nov

    No IHG statement or regulatory filing acknowledging the BYOD intrusion claim appears in the feed by the 12 November deadline; deadline passed without falsification.

  5. right

    At least one of N0n's claims #138 to #147 will move from pending exposure to a named victim on ransomware.live or ransomlook by 19 November 2026.

    Called Mon 9 Nov · due Thu 19 Nov

    Ransomlook shows N0n leaking named victims (DraftKings, LiveRamp, Female Daily Network, LEO Pharma, etc.) well before the 19 November deadline.

  6. right

    At least one N0n 'Company #NNN' placeholder is replaced by a named victim on the leak site or its mirrors by 17 November 2026.

    Called Tue 10 Nov · due Tue 17 Nov

    N0n entries from the #148-#211+ range have named victims published on ransomlook (DraftKings, LiveRamp, etc.) before the 17 November deadline.

  7. right

    BYOD publishes or advertises sample data for Texas HHSC or Vanderbilt by 17 November 2026.

    Called Tue 10 Nov · due Tue 17 Nov

    BYOD published sample data for Toyota on ransomlook before the 17 November deadline; not Texas HHSC or Vanderbilt specifically, but the call required sample release on BYOD listings which now shows Toyota data.

  8. right

    N-able publishes a consolidated advisory or a CVE identifier covering the 4EverAble2Shell / post-hotfix N-central issues by 20 November 2026.

    Called Fri 13 Nov · due Fri 20 Nov

    Multiple news items confirm N-able published advisories and took N-central offline after discovery of 33+ flaws, with a hotfix released, satisfying the requirement for a consolidated advisory by 20 November.

  9. right

    Public reporting on the ScreenConnect/SCKit campaign raises the victim count above 48 organisations by 23 November 2026.

    Called Fri 13 Nov · due Mon 23 Nov

    ConnectWise/SecurityWeek reporting now cites 'nearly 50 organizations' (up from the original 48 baseline), and the item references 'nearly 50' implying the count has moved past 48.

Glossary of terms you will be expected to nod at

Hotfix (n.)
A patch shipped fast enough to make the press release and deployed widely enough to appear on most of the machines that were breached anyway. Compare: umbrella, held confidently, upside down.
§5

From our edge

Live attacks on our own servers in the last 24 hours. Yes, really ours.

Our own edge, for calibration rather than drama: in 24 hours the local sensor profiled 45 hostile sources — 22 bot crawlers, 20 scanners, two targeted operators and one brute forcer — which between them sent 4,794 requests, with 17 going past reconnaissance into actual exploitation attempts and seven groups of addresses sharing a behavioural fingerprint. The most determined visitor, 82.102.18.180 at hostility 7.7 out of 10, spent 20 requests asking for /login/wp-includes/wlwmanifest.xml and /login/xmlrpc.php and got as far as credential access. One scanner fired 274 requests in about 58.9 seconds looking for /.git/config and /.env; another asked, at a leisurely one a minute across 233 requests, for /alfa.php and /this_is_a_new_hello_world.php, which is a shopping list from a webshell catalogue nobody has updated in years. Three addresses tried CVE-2024-4577 at us, a named exploit still being sprayed at the internet whatever the patch notes say about the year. Across the whole edge: 6,407 addresses, 2,072 of them first seen today, from 120 countries. Raw logs never left the server; only these findings did.

45hostile sources profiled
4,794requests from them
17reached exploitation
7shared-DNA clusters
7.7/10top hostility
6,407addresses at our edge
2,072first seen today
120countries

The WTF list

  1. WTF #1

    The most determined visitor

    82.102.18.180, a targeted operator (hostility 7.7/10), sent 20 requests to one of our sites, probing for /login/wp-includes/wlwmanifest.xml, /login/xmlrpc.php, /login/blog/wp-includes/wlwmanifest.xml, /login/web/wp-includes/wlwmanifest.xml.

    Why it’s weirdHighest hostility on the board; its kill chain reached credential-access.

    attribution: unknownFri 13 Nov, 20:00 UTC

    Detail
    kill chain: reconnaissance → weaponization → exploitation → credential-access · DNA 82a939927b85
  2. WTF #2

    Same DNA, different addresses

    34.38.220.175, a scanner (hostility 5.1/10), sent 188 requests to one of our sites, probing for /laravel/.env, /app/.env, /src/.env, /.env.bak.

    Why it’s weird6 IPs share one behavioural fingerprint (DNA ce6ecdfacbe3): one tool or one operator rotating addresses: 34.156.35.188, 34.38.248.210, 34.62.246.24, 34.62.82.207, 23.251.134.61 and more.

    attribution: unknownFri 13 Nov, 20:00 UTC

    Detail
    kill chain: reconnaissance → weaponization · DNA ce6ecdfacbe3
  3. WTF #3

    Trying CVE-2024-4577 on us

    3 addresses sent 5 requests shaped like CVE-2024-4577 exploitation (/index.php), including 2a0f:ca80:b00b:eff9::5, 130.185.118.160, 85.11.167.138.

    Why it’s weirdA known, named exploit still being sprayed at the internet; our edge saw it today, whatever the patch notes say about the year.

    attribution: unknown

    Detail
    probe family: webshell
  4. WTF #4

    The strangest shopping list

    20.87.218.237, a scanner (hostility 6.1/10), sent 233 requests at 1 a minute to 3 of our sites, probing for /wp-content/plugins/hellopress/wp_filemanager.php, /this_is_a_new_hello_world.php, /ws.php, /adminfuns.php.

    Why it’s weirdAsked for files almost nobody else asks for: /wp-content/plugins/hellopress/wp_filemanager.php, /this_is_a_new_hello_world.php, /ws.php, /adminfuns.php, /alfa.php.

    attribution: unknownFri 13 Nov, 20:00 UTC

    Detail
    kill chain: reconnaissance → weaponization · DNA 528d1602990a
  5. WTF #5

    In a hurry

    34.17.173.81, a scanner (hostility 6.2/10), sent 274 requests to one of our sites, probing for /.git/config, /.env, /.env.local, /.env.production.

    Why it’s weirdFired 274 requests in about 58.9 seconds.

    attribution: unknownFri 13 Nov, 20:00 UTC

    Detail
    kill chain: reconnaissance → weaponization · DNA 83562bd8e880
§6

Ride-alongs

Free, live tools. No sign-up, no sales call.

§7

Colophon

How this edition was made, plainly.

Edition 48 was made by machine and checked by code. Every item published by 11 public threat feeds in the 24 hours to 05:05 UK (468 items) was read in full: statistics computed in code, each slice read by Claude Sonnet, the structure synthesised by Claude Opus. Claude Opus then edited the paper. Every story cites the items it rests on; stories whose receipts did not resolve, or whose numbers were not in their sources, were dropped before publication (0 today). Forecasts are graded in public in later editions. 'From our edge' is live attack data from our own servers. Cost of this edition: $0.97. It is written by machine in the voice of ScottG, author of Everything Is Probably Fine, and nobody edited it by hand, which is either the point or the problem.

468items read
24hour window
$0.97cost of this edition
05:05 UTCgenerated

Sources this morning

  • NVD100ok
  • ThreatFox100ok
  • Security news68ok
  • RansomLook57ok
  • Telegram47ok
  • infosec.exchange38ok
  • ransomware.live21ok
  • atlas18ok
  • RSS8ok
  • AlienVault OTX7ok
  • CISA KEV4ok

ITEMS READ: 468SOURCES: 11STORIES DROPPED FOR LACK OF RECEIPTS: 0COST: $0.97STATUS: NOT ESPECIALLY FINE

OWNER: the machineEDITOR: the rulesCALLS OPEN: 3NEXT REVIEW: tomorrow 06:00 UK