Lead · Vulnerabilities · INST-206

Four vulnerabilities joined the exploited list, and every score said not to bother

CISA added Apache Roller, VMware vCenter, Cisco Expressway and SAP NetWeaver to the Known Exploited Vulnerabilities catalogue on Thursday, and the highest exploit-prediction score of the four was 0.01191.

confidence 0.95

If you had spent this week ranking your patch queue by probability, you would have skipped all four of them: Apache Roller at 0.01191, VMware vCenter at 0.00843, Cisco Expressway at 0.00417, SAP NetWeaver at 0.00276. Not one of them crosses one and a half per cent. All four are on the list that exists because someone, somewhere, already did it.

The vCenter entry, CVE-2026-123890, is an authentication bypass in the VMware Directory Service that lets an unauthenticated attacker with network access to vCenter reach it anyway; SecurityWeek and Dark Reading both report that the suspected PRC-nexus group Armory Castle, formerly tracked as UNC5221, was exploiting it as a zero-day in early September 2026, at CVSS 9.8, to take admin access and steal data. The Roller entry, CVE-2026-129448, is deserialisation of untrusted data at the XML-RPC endpoint, which accepts and deserialises vendor extension types before authentication, which is to say it runs your code and then asks who you are, like a nightclub that searches people on the way out.

And CVE-2026-76466, hard-coded credentials in Cisco Expressway, allows a remote attacker with persistent cookies to execute commands as root on a clustered device via the database service; The Hacker News reports it was used against a Russian organisation shortly after a proof of concept went public, breaching four devices and reaching the internal network behind them. The gap between a PoC landing and a KEV entry is now shorter than most change windows. That is the whole story.

Why it mattersToday's order is vCenter, Expressway, Roller, NetWeaver, and your EPSS-sorted backlog is not a plan, it is a sentiment.

Today’s revelation

Every exploit-probability score is a forecast of rain; the KEV is a photograph of you, already wet.

— Sturgeon’s Revelations · Probably Fine Daily No. 55

§2

Top stories

The rest of the last 24 hours, with receipts.

Ransomware

N0n has stopped encrypting things and is doing better than ever

Ninja Signal counted 369 leak-site claims in the last seven days from 50 of the 281 groups it tracks, and N0n posted 92 of them against 36.81 expected from its own prior rate, p below 0.001. AlienVault describes the group, which emerged in September 2026, as doing data theft and extortion without traditional encryption at all. The week's listings include Halfords Group, Guardant Health, where Hackmanac records a claim of 162,870 records, and GRAIL, where darkwebsonar logged a claim of 56,684 healthcare provider records and 73 N0n listings in seven days, while JPMorgan Chase, Visa and Mastercard have all pointed at third-party service providers after N0n leaked data allegedly tied to their customers. On Darkfeed's 2026 table N0n sits ninth for the year with 198, a figure it is currently rewriting weekly.

Why it mattersRestore-from-backup is not a control against a group whose only product is a copy, and your third parties are the ones holding the copy.

confidence 0.85

AI & security

148 actions, sixteen hours, one real bug: the autonomous pentest framework turns up in an actual intrusion

Amazon's threat intelligence team reports that Chinese-speaking operators pointed Cairn, an open-source autonomous AI penetration-testing framework, at two East Asian universities; the agents executed 148 actions over sixteen hours and the single success was a previously unknown path-traversal flaw in an open-source medical imaging application, then reused against the second university. In the same 24 hours Bell Canada confirmed an "agentic" threat actor breached its systems on 17 November and took enterprise customer data, including from a system used to order and provision telecommunications circuits, and the European Commission confirmed an AI-assisted attack that it contained by pulling services offline, with no evidence of a data breach. Meanwhile an a16z survey of CISOs at organisations above 1,000 staff has 36% comfortable letting AI agents act autonomously on low-risk security actions today, 60% expecting to be within twelve months, and 7% expecting to trust AI with most actions.

Why it mattersThe interesting number is not the zero-day, it is the sixteen hours of patient, cheap, unsupervised noise that preceded it, and whether any of it reached your console.

confidence 0.88

Threat actors

The outage was the incident: N-able breach takes Mail Assure down worldwide

N-able says attackers who breached one of its servers caused a major outage affecting Mail Assure and SpamExperts Hosted Cloud services worldwide, and SecurityWeek notes the incident could have led to credential theft. On the same morning J.P. Morgan Chase warned customers that a cyberattack at a third-party service provider exposed personal information and marketing data, with Visa and Mastercard citing third-party incidents too. Two different shapes of the same dependency: one you notice because the mail stops, one you notice because a bank writes to you.

Why it mattersYou cannot patch a supplier, so the only control you have left is knowing, today, which of your services would simply stop.

confidence 0.90

Vulnerabilities

A guest virtual machine read about 127 GB of the landlord's memory

Researchers used CVE-2026-99312, a flaw in KVM, the hypervisor built into the Linux kernel, to read roughly 127 GB of a cloud server's memory from inside a virtual machine of their own on DigitalOcean; the flaw affects AMD hosts with nested virtualisation enabled, and on a shared server that memory can include other customers' machines. The kernel fix is three lines of prose about reality: unconditionally reset the MMU context when emulating INIT, because the page role includes guest_mode, so an INIT while L2 runs with CR0.PG=0 leaves KVM running L1 with a stale context. The gap between a stale context and 127 GB is the entire commercial premise of public cloud.

Why it mattersIf you run nested virtualisation on AMD, tenancy isolation is now a patch level rather than an architecture.

confidence 0.85

Ransomware

A loud night on the leak sites: a Boeing subsidiary, an Italian energy major, a small-town police force

INC Ransom posted a claim against Boeing Sikorsky Aircraft Support, a Boeing subsidiary at Fort Campbell, Kentucky, and another against eni.it itemising 7,105 contract files at 18.78 GB and 28,562 finance and accounting files at 1.91 GB, with Darkfeed trailing it as a major aerospace breach. Everest put eight names up inside three hours, and ransomware.live carried each of these one to three hours before ransomlook did. Every figure here is the attacker's own, posted by people whose business model rewards a big number.

Why it mattersTreat leak-site inventories as marketing with a file count, then check whether any of those file counts describe data you sent someone.

confidence 0.80

Vulnerabilities

The no-patch pile, plus one vendor that has stopped pretending

SonicWall told every customer still running Generation 6 firewalls to retire and replace them immediately and stop exposing them at the internet edge, because no further security patches are coming even for critical flaws, and cited attackers using automation and AI to find unsupported edge devices. A newly disclosed and unpatched Windows technique called FastUnlock lets someone physically holding a locked machine start programs without signing in, with a limited account rather than administrator control. CVE-2026-19793 in WP Import Export Lite, all versions to 3.9.32, lets unauthenticated comment content become PHP execution when an administrator later runs an export or import, with no patch and the advice being to review pending comments. And the NGINX2Root set of five disclosed flaws includes CVE-2026-123007, an HTTP/2 issue reported at CVSS 9.2 affecting unpatched releases dating back to 1.9.5.

Why it mattersEnd-of-life is a risk acceptance your supplier signed on your behalf, and the renewal date is the only thing that will tell you.

confidence 0.80

Oddities

The advanced cryptographic tools turned out to be a bank transfer

The chief technology officer and co-founder of ransomwarehelp.com, a six-year-old Florida business that advertised recovery of ransomed data using advanced cryptographic tools, has been charged with wire fraud after an FBI sting found the company was simply paying the attackers for a decryption key and charging a markup. Six years. A website. A category of vendor that exists because the alternative conversation is unbearable.

Why it mattersIf a recovery supplier will not describe the mechanism in a sentence you can repeat to a regulator, you are buying a negotiation with a logo on it.

confidence 0.85
§3

Hidden

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

The two biggest groups of the year are cooling, a newcomer is surging, and two trackers you treat as independent are reading from the same page an hour apart.

  1. Exception 01

    The league leaders are idling while the newcomer runs

    Qilin (1,194 claims in 2026) and The Gentleman (1,047) dominate Darkfeed's year table, yet every Atlas signal carried for both in this window is flagged cooling, across manufacturing, professional services, transport and four countries apiece, while N0n, ninth on the same table at 198 for the year, posted 92 claims in seven days against 36.81 expected.

    confidence 0.70
    Wrong if
    Qilin or The Gentleman post 25 or more fresh leak-site claims in the seven days to 28 November, which would make 'cooling' an artefact of the window rather than a trend.
    Watch
    Weekly claim counts for Qilin, thegentlemen and N0n on ransomware.live and ransomlook.
  2. Exception 02

    Your two ransomware trackers are one tracker with a delay

    Across this window ransomware.live consistently precedes ransomlook on the same victim by roughly one to three hours: INC Ransom on Boeing by 1.6 hours, Everstake by 1.4, Dark Project on VanMor by 1.4, akira on Pecora by 1.0, aurora on William Cook by 1.0. Seeing a claim twice is not corroboration, it is the second site finishing its scrape.

    confidence 0.75
    Wrong if
    A week in which ransomlook carries three or more victim claims that never appear on ransomware.live, or leads it by more than an hour.
    Watch
    Timestamp deltas on the next ten shared victim listings.
  3. Exception 03

    If you learn about KEV from your timeline, you are a day late

    CVE-2026-129448 was published to the KEV catalogue on 20 November and the first Mastodon post explaining it to defenders arrived 28.4 hours later, on 21 November at 04:25. The KEV feed is free, machine-readable and faster than the people summarising it, including us.

    confidence 0.80
    Wrong if
    The next three KEV additions appear in community feeds within six hours of publication.
    Watch
    Time from cisa-kev entry to first community post for the next KEV batch.

Nothing to seeNo new KEV entries dated today; Thursday's batch of four is still the batch. The NVD feed in this window stops at 16:44 on Wednesday after 100 items, which is a weekend and not a ceasefire. And despite roughly a hundred WordPress plugin and theme CVEs published on Wednesday, nothing in the last 24 hours ties any of them to exploitation in the wild, which is not the same as nobody trying.

Deep read: 569 items read

§4

Forecast

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

Today’s calls

  1. DueSat 28 Nov7d

    N0n posts at least 40 further leak-site claims in the seven days to 28 November, continuing its surge above the 36.81 expected from its own prior rate.

    Wrong ifFewer than 40 N0n claims recorded across ransomware.live and ransomlook in that window.

    f-2026-11-21-1

    Confidence0.72
  2. DueSat 28 Nov7d

    Rhysida keeps running hot and posts at least five more victim claims in the seven days to 28 November, against 1.86 expected from its own prior rate.

    Wrong ifFour or fewer rhysida claims recorded in that window.

    f-2026-11-21-2

    Confidence0.60
  3. DueFri 4 Dec13d

    CVE-2026-123007, the NGINX HTTP/2 flaw in the NGINX2Root set, is added to the CISA KEV catalogue on or before 4 December.

    Wrong ifThe CVE is not present in the KEV catalogue on the morning of 5 December.

    f-2026-11-21-3

    Confidence0.30

Scorecard

2 right · 1 wrong

  1. wrong

    CISA adds the SSHole OpenSSH flaw, CVE-2026-123158, to the KEV catalogue.

    Called Wed 11 Nov · due Sat 21 Nov

    Due date is today (21 Nov). CVE-2026-123158 (SSHole/OpenSSH) does not appear in any KEV addition in today's items; only Apache Roller, VMware vCenter, SAP NetWeaver, and Cisco Expressway were added on 20 Nov.

  2. right

    CISA adds Apache Roller CVE-2026-129448 to the KEV catalogue by 26 November.

    Called Thu 19 Nov · due Thu 26 Nov

    CISA KEV added CVE-2026-129448 (Apache Roller) on 20 Nov, before the 26 Nov deadline.

  3. right

    The Apache Roller flaw CVE-2026-129448 reaches the KEV catalogue by 27 November.

    Called Fri 20 Nov · due Fri 27 Nov

    CISA KEV added CVE-2026-129448 (Apache Roller) on 20 Nov, before the 27 Nov deadline.

Glossary of terms you will be expected to nod at

Agentic (n.)
Of an intrusion: conducted at a tempo that implies nobody stopped for lunch. Of a product: the same word, printed on the box, by people who have also not stopped for lunch.
§5

From our edge

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

Our own edge profiled 100 hostile sources in the last 24 hours, 50 bot crawlers, 47 scanners, two brute forcers and one targeted operator, sending 9,386 requests between them; 55 went past reconnaissance into exploitation attempts and 12 groups of addresses shared a behavioural fingerprint. Sixteen separate IPs share DNA ce6ecdfacbe3 and spend their day asking for /.env, /api/.env, /backend/.env and /app/.env, which is one operator with a rotation policy and a shopping list. One address, 195.178.110.48, fired 212 requests in about 1.1 seconds, which is not a person. Another, 84.17.60.236, asked twenty times for //wp-includes/wlwmanifest.xml and its variants, a file almost nobody else requests, and 172.212.145.132 sent 170 requests hunting /admin.php, /403.php, /css.php and /cong.php at hostility 6.8 out of 10, the highest on the board. A single address, 195.178.110.31, sent exactly one request shaped like CVE-2024-4577 exploitation against /index.php, which tells you precisely how much the internet cares what year the patch notes say. Across the window we saw 8,993 addresses from 113 countries, 3,333 of them new today. Raw logs never left the server.

100hostile sources profiled
9,386requests from them
55reached exploitation
12shared-DNA clusters
8.2/10top hostility
8,993addresses at our edge
3,333first seen today
113countries

The WTF list

  1. WTF #1

    The most determined visitor

    172.212.145.132, a scanner (hostility 6.8/10), sent 170 requests to one of our sites, probing for /admin.php, /403.php, /css.php, /cong.php.

    Why it’s weirdHighest hostility on the board; its kill chain reached weaponization.

    attribution: WordPress reconnaissance scanner (sensor's best match, 45%)Fri 20 Nov, 16:00 UTC

    Detail
    kill chain: reconnaissance → weaponization · DNA dfef4501b216
  2. WTF #2

    Same DNA, different addresses

    34.76.86.40, a scanner (hostility 4.7/10), sent 317 requests to one of our sites, probing for /api/.env, /backend/.env, /.env, /app/.env.

    Why it’s weird16 IPs share one behavioural fingerprint (DNA ce6ecdfacbe3): one tool or one operator rotating addresses: 34.62.178.133, 91.148.244.131, 207.175.81.69, 34.77.76.113, 34.79.189.227 and more.

    attribution: unknownFri 20 Nov, 16:00 UTC

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

    Trying CVE-2024-4577 on us

    1 address sent 1 requests shaped like CVE-2024-4577 exploitation (/index.php), including 195.178.110.31.

    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

    84.17.60.236, a scanner (hostility 6.3/10), sent 20 requests to one of our sites, probing for //wp-includes/wlwmanifest.xml, //xmlrpc.php, //blog/wp-includes/wlwmanifest.xml, //web/wp-includes/wlwmanifest.xml.

    Why it’s weirdAsked for files almost nobody else asks for: //wp-includes/wlwmanifest.xml, //xmlrpc.php, //blog/wp-includes/wlwmanifest.xml, //web/wp-includes/wlwmanifest.xml, //wordpress/wp-includes/wlwmanifest.xml.

    attribution: unknownFri 20 Nov, 16:00 UTC

    Detail
    kill chain: reconnaissance → weaponization · DNA dfef4501b216
  5. WTF #5

    In a hurry

    195.178.110.48, a scanner (hostility 4.7/10), sent 212 requests to one of our sites, probing for /.env.local, /.env.development, /.env, /.env.bak.

    Why it’s weirdFired 212 requests in about 1.1 seconds.

    attribution: unknownSat 21 Nov, 03:00 UTC

    Detail
    kill chain: reconnaissance → weaponization · DNA bf298923109a
  6. WTF #6

    Somebody we might know

    157.230.254.35, a scanner (hostility 5.9/10), sent 24 requests at 2 a minute to 2 of our sites, probing for //images/images/cache.php, //cgi-bin/cgi-bin/cgi-bin/cgi-bin/cache.php, //images/images/images/images/images/images/images/images/cache.php, //images/images/images/images/images/images/images/cache.php.

    Why it’s weirdThe sensor's most confident campaign match of the day: Path-normalization bypass scanner (custom/unknown variant) at 58%.

    attribution: Path-normalization bypass scanner (custom/unknown variant) (sensor's best match, 58%)Sat 21 Nov, 05:00 UTC

    Detail
    kill chain: reconnaissance → weaponization · DNA bccc30d19899
§6

Tempo

Who is posting faster than their own normal. Computed from the Signal graph, not written by the editor.

369claims, 7 days
50of 281 groups active
N0n92 this week · surging
PixelLeaks31 this week · new
SoulGroup26 this week · new
  • SurgingN0n — 92 claims in 7 days against 36.8 expected from its own prior rate (p < 0.001)
  • Surgingrhysida — 11 claims in 7 days against 1.9 expected from its own prior rate (p < 0.001)
  • SurgingShadowByt3$ — 5 claims in 7 days against 0.7 expected from its own prior rate (p < 0.001)
  • Surgingaurora — 7 claims in 7 days against 1.9 expected from its own prior rate (p = 0.004)
  • Surgingsettra — 11 claims in 7 days against 5.5 expected from its own prior rate (p = 0.027)
  • Backblackout — first claim after 124 days of silence; 1 this week

Gone quiet: titan (60 days), moneymessage (61 days), unsafe (61 days), killsec (64 days).

Sector pressure, 30 days

  • UpRetail & E-Commerce — 1.3× its own prior share (109 claims, 84 expected)
  • UpHealthcare — 1.2× its own prior share (152 claims, 123 expected)
  • DownAgriculture and Food Production — 0.4× its own prior share (17 claims, 46 expected)
  • DownTransportation — 0.7× its own prior share (39 claims, 57 expected)

26% of claims carry no sector label.

Brand churn

Of the 243 leak-site brands first seen since 2021-09-30, 75% were still claiming 30 days after their first post and 70% were still claiming 60 days after their first post and 67% were still claiming 90 days after their first post and 61% were still claiming 120 days after their first post (Kaplan-Meier; quiet = 60 days without a claim). 18% of the 218 old enough to judge posted on one day and never again; 38% made at least half their claims in their first week. New brands this month: 12 — newest: SoulGroup (2026-11-19, 26 claims), SilentVault (2026-11-19, 8 claims), Handsome Boy Hacking Club (2026-11-19, 4 claims).

Brands born before coverage began are excluded (their true age is unknown); a rebrand counts as one brand going quiet and another being born.

Targeting

Over the last 180 days, among brands with at least 20 labelled claims: AuditTeam put 32% of its claims on Russia (168.8× every other brand's share); safepay put 33% of its claims on Germany (6.3× every other brand's share); SilentRansomGroup put 100% of its claims on Professional Services (6.4× every other brand's share). Widest spread: nightspire across 10.6 effective sectors.

Lift = the brand's share of a sector or country over every other brand's share in the same window; exact binomial tails, Bonferroni-adjusted for the categories tested. Claims with no sector or an Unknown/Other label are excluded.

Posting clock

Over the last 180 days, among brands with at least 30 claims: clop released 94% of its claims on Wednesdays, half of them in 13:00–15:59 UTC; Deadlock released 60% of its claims on Fridays, half of them in 12:00–13:59 UTC; akira posted only 2 of 196 claims at weekends and half its claims in 12:00–13:59 UTC; SilentRansomGroup posted only 2 of 92 claims at weekends and half its claims in 19:00–22:59 UTC. Leak sites overall are 40% quieter per weekend day than per weekday.

all brands by hour of day, 00–23 UTC; half of all claims land in 09:00–17:59 UTC.

Times are when trackers first saw each claim on the leak site, in UTC: not the attack time, not the operator's clock, and a site that is only reachable at certain hours shows the same pattern; batch releases show as spikes. Hour-of-day concentration: Rayleigh test; weekday shares: exact binomial against every other brand, Bonferroni-adjusted.

Leak-site claims as discovered by trackers, not confirmed breaches. Surprise compares the last 7 days with the group's own prior 173 days (Poisson). Victim claims: ransomware.live and leak-site trackers

Today’s numbers as JSON →

§7

Ride-alongs

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

§8

Colophon

How this edition was made, plainly.

Edition 55 was made by machine and checked by code. Every item published by 15 public threat feeds in the 24 hours to 05:05 UK (569 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: $1.16. 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.

569items read
24hour window
$1.16cost of this edition
05:05 UTCgenerated

Sources this morning

  • Signal Atlas110ok
  • NVD100ok
  • ThreatFox100ok
  • Telegram72ok
  • RansomLook44ok
  • infosec.exchange37ok
  • Security news35ok
  • ransomware.live23ok
  • atlas15ok
  • AlienVault OTX12ok
  • RSS8ok
  • signal:kev4ok
  • CISA KEV4ok
  • signal:ioc-family4ok
  • Ransomware tempo · Ninja Signal graph1ok

ITEMS READ: 569SOURCES: 15STORIES DROPPED FOR LACK OF RECEIPTS: 0COST: $1.16STATUS: NOT ENTIRELY FINE

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