Koffei

Daily CyberNews · 17/11/2026

‹ 16/11/202618/11/2026 ›
Today in 30 seconds
  • 🔴 Hijacked Optimizely snippets served a skimmer, clipper and iOS exploit kits on 178 sites: hunt for the IOCs.
  • 🔴 Patch LiteLLM to 1.83.7 and Starlette to 1.0.1: scans for the KEV-listed RCE chain have resumed.
  • 🔴 Update TYPO3: CVE-2026-125160 is already exploited; also serve the upload folder with a restrictive CSP.
  • 🔴 Take GeoVision GV-VMS off the internet and upgrade to V17.4.8.58: exposed servers were hijacked for mining.
  • 🟠 Disable the SRP mechanism in Cyrus SASL: CVE-2026-129483 allows login as any user with a public PoC.

Vulnerabilities 14

🟠 High impact

NoisyTCP: pre-auth remote kernel crash in Linux TCP, almost every distribution affected

CVE-2026-99288CVSS 7.5Linux TCPNoisyTCP

WhatA race in Linux TCP socket lookup lets an unauthenticated remote attacker crash the kernel through any plain TCP listener; no exploitation or public exploit.

DoInstall a fixed kernel such as 7.3.4, 6.18.63 or 6.12.117, or as a stopgap set net.ipv4.tcp_syncookies=0 in every namespace.

Details and sources
CVE-2026-99288 (CVSS 3.1 7.5 from Red Hat, AV:N/AC:H; NVD has no score yet; CWE-476; EPSS under 1%; not in KEV) is a race in the lockless TCP established-socket lookup: a request socket freed and reused under SLAB_TYPESAFE_BY_RCU can be returned unhashed with a NULL listener, and dereferencing it panics the kernel. It is reachable without login against any plain TCP listener (default Nginx, Apache or SSH) on hosts with two or more RX queues or RPS; in a container the whole host goes down. A second path can give a remote use-after-free but is hard to trigger. The researcher's tuned PoC crashed QEMU/KVM guests in 5 seconds to 2 minutes, but needed about 14 hours on a quiet 8-queue bare-metal server. The bug dates from a 2015 commit (079096f103fa); the fix is mainline commit 4e2b3ef19e0e, shipped in stable 7.3.4, 6.18.63 and 6.12.117 (Debian trixie 6.12.117-1, sid 7.3.4-1). Older LTS branches (6.6, 6.1, 5.15, 5.10) and Ubuntu were still listed as unfixed by one secondary source. Stopgaps: net.ipv4.tcp_syncookies=0 in every network namespace (weakens SYN-flood protection), or a single RX queue with RPS/RFS off. No exploitation in the wild and no public exploit. Openwall (source) · OpenCVE (source)
🔴 Exploited or active

Fresh scans for the LiteLLM and Starlette pre-auth RCE chain

CVE-2026-42271CVE-2026-48710in KEVLiteLLMStarlette

WhatHoneypots saw new scans for the LiteLLM and Starlette pre-auth RCE chain, using malformed Host headers to slip past filters.

DoUpgrade LiteLLM to 1.83.7 and Starlette to 1.0.1, and filter on invalid hostname characters rather than exploit strings.

Details and sources
SANS ISC honeypots began receiving POST /mcp-rest/test/tools/list requests over the weekend with a JSON body setting the MCP transport to stdio and the command to id, Authorization: Bearer nothing, User-Agent Python/3.14 aiohttp/3.14.5 and malformed Host headers such as localhost] [localhost 127.0.0.1] litellm.proxy, apparently a published PoC tweaked to slip past filters. The chain pairs CVE-2026-48710 (Starlette "BadHost", CVSS 3.1 6.5, no Host header validation lets request.url checks be bypassed; before 1.0.1, fixed in 1.0.1; in KEV since 2 September; EPSS about 70%) with CVE-2026-42271 (LiteLLM, CVSS 4.0 8.7 / 3.1 8.8, the MCP test endpoints spawn any stdio command for any API key; 1.74.2 to before 1.83.7, fixed in 1.83.7; Red Hat OpenShift AI 2.25, 3.3 and 3.4 too; in KEV since 8 June; EPSS about 91%). Patch both, and filter on invalid hostname characters rather than on exploit strings. SANS ISC (source) · OpenCVE (source)
🔴 Exploited or active

Exposed GeoVision GV-VMS servers hijacked for cryptomining at 13 organisations

GeoVision GV-VMSXMRigBYOVD

WhatAttackers hijacked exposed GeoVision GV-VMS servers at 13 organisations, tampered with antivirus via a vulnerable driver and installed XMRig.

DoRemove internet exposure, upgrade to V17.4.8.58, and hunt for shells or PowerShell spawned by the GeoVision service.

Details and sources
Huntress traced intrusions, mostly on 8-10 November, that each began with a GeoVision service on Windows hosts running GV-VMS 17.4.8.0 (build of 24 April 2024); the fixed build is V17.4.8.58. Huntress has not identified the flaw used; the likely candidates are the three unauthenticated RCE flaws in the WebCam Server component from GeoVision's April 2026 advisory (no CVE confirmed for these attacks). The attackers loaded a driver known from BYOVD attacks to tamper with antivirus, set three IFEO debugger hijacks pointing security processes to systray.exe, and ran a loader (every 15 minutes and at boot) that installs XMRig and finds its servers through BitTorrent Mainline DHT and a PKARR resolver. Advice: remove internet exposure (VPN or IP allow-list) and hunt for shells or PowerShell spawned by the GeoVision service. Help Net Security (source)
🟠 High impact

Apache Shiro LDAP realm accepts any password for cached users

CVE-2026-129349CVSS 9.1Apache ShiroLDAP

WhatWith LDAP realm authentication caching enabled, Apache Shiro accepts any password for users who logged in once; a PoC is reported.

DoUpgrade shiro-core to 3.0.1, or turn off authentication caching or set SimpleCredentialsMatcher.

Details and sources
CVE-2026-129349 (CVSS 4.0 9.1, CWE-287, EPSS under 1%, not in KEV), published 15 November: with DefaultLdapRealm or JndiLdapRealm and authentication caching enabled (not the default), a user who has logged in once can then be logged in with any password, because the cached path skips the LDAP bind and the default AllowAllCredentialsMatcher accepts everything. Anyone who knows a valid username can take over that account. shiro-core before 3.0.1 is affected; 3.0.1 stops installing AllowAllCredentialsMatcher by default. Workarounds: turn off authentication caching or set SimpleCredentialsMatcher. A PoC is reported; no exploitation. OpenCVE (source) · CVE Brief (source)
🟠 High impact

Cyrus SASL SRP bypass lets anyone log in as any user, with a public PoC

CVE-2026-129483CVSS 9.3Cyrus SASLpublic PoC

WhatA Cyrus SASL SRP flaw lets a remote attacker log in as any existing user without the password; a public PoC exists.

DoDisable the SRP mechanism until a patched Cyrus SASL package ships.

Details and sources
CVE-2026-129483 (CVSS 3.1 9.1 / 4.0 9.3, CWE-287, EPSS under 1%, not in KEV), published 15 November: srp_server_mech_step2 does not reject a client public value A that is a multiple of N, so sending A = N forces a zero shared secret and gives the attacker a session as any existing user without the password, over the network and without authentication. Cyrus SASL up to 2.1.28 is affected; the fix is commit fb3724f, with no fixed release named yet, and a proof of concept is on a public GitHub gist. Only services offering the SRP mechanism are exposed: disable SRP until a patched package ships. No exploitation reported. OpenCVE (source) · CVE Brief (source)
🔴 Exploited or active

Trend Micro Apex One: 18 agent flaws fixed, plus a PHP update against PHP2Shell

Added after publishingTrend Micro Apex OneCVE-2026-109618CVE-2026-109619in KEV

WhatTrend Micro fixed 17 local privilege escalations in the Apex One agent and updated the server's bundled PHP for the exploited PHP2Shell flaws.

DoInstall Apex One SP1 CP Build 20113 on-prem, or agent 14.0.21067 for Apex One as a Service and SEP.

Earlier coverage: 16/1113/11
Follow-ups: 19/11
Details and sources
CERT-FR advisory CERTFR-2026-AVI-1542 (16/11/2026) relays Trend Micro's bulletin KA-0024778 of 13 November for Apex One (on-prem) 2019, Apex One as a Service and the Vision One Endpoint Security Standard Endpoint Protection (SEP) agent. Seventeen flaws let a local user who can already run low-privileged code reach SYSTEM (CVSS 3.1 7.8 each): insecure permissions on the agent temp directory CVE-2026-111773, DLL search-order hijacking CVE-2026-111774, CVE-2026-111777, an unprivileged OIPC interface CVE-2026-111775, COM hijacking CVE-2026-111778, origin validation errors CVE-2026-111779, CVE-2026-111780, CVE-2026-111781, CVE-2026-111789, CVE-2026-111790, stack and heap buffer overflows CVE-2026-111782, CVE-2026-111783, CVE-2026-111785, a null pointer dereference CVE-2026-111784, an out-of-bounds read CVE-2026-111786 and type confusion CVE-2026-111787, CVE-2026-111788; most came in through ZDI. CVE-2026-111776 (CVSS 5.5) is a heap overflow that crashes the agent service. The server's bundled PHP is also updated for the PHP2Shell flaws CVE-2026-109618 and CVE-2026-109619 (CVSS 4.0 9.2, exploited and in KEV), detailed in KA-0024802. Fixed in Apex One SP1 CP Build 20113 (agent 14.0.0.20113) and agent 14.0.21067 for Apex One as a Service and SEP (October 2026 maintenance release). No exploitation of the Apex One flaws is reported; their CVE records are not yet published, so there is no EPSS yet. CERT-FR (source) · Trend Micro (source)
⚪ Context

Apple narrows macOS app permissions with AI agents in mind

Added after publishingApplemacOS 27.3 betaapp permissions

WhatApple is splitting and tightening macOS permissions such as Full Disk Access and Accessibility, framed around AI-agent risks; no release date.

DoReview which apps hold Full Disk Access or Accessibility rights.

Details and sources
previewed in macOS 27.3 beta 2 and announced on 16 November, the "Device Control and Data Access" setting is split into System & App Protection Override and Accessibility, and Full Disk Access is tightened. Users pick the requesting app from a list, are pushed to grant access only to specific apps or folders, can set an app to ask every time, and see an indicator when one app controls another. Existing grants stay; backup and assistive-technology apps can request entitlements to keep their behaviour on new installs. Apple frames it as informed consent before "extraordinary access" that "puts all of the data on their Mac at risk", and The Hacker News ties it to AI-agent risks. No release date and no CVE. A good moment to review which apps hold Full Disk Access or Accessibility rights. The Hacker News (source)
🟡 Worth knowing

iPhone flaw let apps record the microphone after permission was denied

Added after publishingCVE-2026-65339CVSS 5.0iOSmicrophone

WhatA fixed iPhone flaw let an installed app record the microphone after permission was denied, even in the background; no exploitation reported.

DoMake sure devices run iOS 26.6.1 or 18.7.11, macOS Tahoe 26.6.2 or Sequoia 15.8 or later.

Details and sources
CVE-2026-65339 (CVSS 3.1 5.0 from CISA ADP, AV:L/AC:L/PR:L/UI:R, Apple's record has no score; CWE-693; EPSS under 1%; not in KEV, SSVC exploitation none) let an installed app send a forged audit token naming the audiomxd audio service, which then treated the app as itself and skipped the TCC consent check: microphone audio could be captured after the user said no, in the background and with the device locked, and on one test build the orange microphone indicator stayed off. Only microphone capture through that service was affected. Found by Nik Tsytsarkin (Meta Red Team X) and, separately, Mustafa Calap. Apple's August advisory only said "a logic issue was addressed with improved checks"; fixed in iOS and iPadOS 26.6.1 (17/08/2026) and 18.7.11 (26/10/2026), macOS Tahoe 26.6.2, macOS Sequoia 15.8, and tvOS, watchOS and visionOS 27. No exploitation reported. The Hacker News (source) · OpenCVE (source)
🟠 High impact

Tenable Identity Exposure 3.93.7 and 3.77.19 fix 52 flaws in bundled components

Added after publishingTenable Identity Exposure52 CVEsCVE-2026-8925CVSS 9.8

WhatTenable fixed 52 flaws in bundled curl, OpenSSL, 7-Zip, node-forge and SQL Server components of Identity Exposure; no exploitation mentioned.

DoUpgrade Tenable Identity Exposure to 3.93.7 or 3.77.19.

Details and sources
CERT-FR advisory CERTFR-2026-AVI-1550 (17/11/2026) relays Tenable's TNS-2026-35 (16/11/2026, overall risk Critical) for Identity Exposure 3.93.6 and earlier and 3.77.18 and earlier. All 52 CVEs sit in third-party components: curl, updated to 8.23.0 (CVE-2026-19931, CVE-2026-11856, CVE-2026-9079, CVE-2026-8925 at CVSS 9.8, CVE-2026-8927 at 9.1, plus CVE-2026-92392, CVE-2026-5545, CVE-2026-6253, CVE-2026-6429, CVE-2026-7168, CVE-2026-7009); OpenSSL, updated to 4.0.3 (CVE-2026-84782, CVE-2026-84783, CVE-2026-18798, CVE-2026-63072, CVE-2026-63075, all High); 7-Zip, updated to 26.05 (CVE-2026-48095, CVE-2026-48092, CVE-2026-14266, CVE-2025-11001, CVE-2025-11002, CVE-2022-29072, CVE-2023-31102, CVE-2025-53816, CVE-2025-53817, CVE-2026-48103, CVE-2026-48111, CVE-2026-48101, CVE-2026-48112, CVE-2026-48102, CVE-2026-48104); node-forge 1.4.3 (CVE-2026-118779); and Microsoft SQL Server 2019 through KB5128494 (CVE-2026-65674, CVE-2026-94512, CVE-2026-67625, CVE-2026-68883, CVE-2026-85886, CVE-2026-96956, CVE-2026-85888, CVE-2026-88081, CVE-2026-85876, CVE-2026-85919, CVE-2026-88082, CVE-2026-88088, CVE-2026-32176, CVE-2026-70189, CVE-2026-88089, CVE-2026-88085, CVE-2026-88084, CVE-2026-88086, CVE-2026-67628, CVE-2026-94520). No exploitation is mentioned. CERT-FR (source) · Tenable (source)
🟠 High impact

Juniper Junos OS Evolved inherits the Linux IPv4 TCP use-after-free, fix still pending

Added after publishingCVE-2026-98432CVSS 9.8Juniper Junos OS Evolved

WhatEvery Junos OS Evolved version inherits a Linux TCP use-after-free with possible RCE or DoS, and Juniper's fix is still pending.

DoApply the workaround in Juniper bulletin JSA119568 until the fix is released.

Earlier coverage: 04/1103/11
Details and sources
CERT-FR advisory CERTFR-2026-AVI-1551 (17/11/2026) relays Juniper's out-of-cycle bulletin JSA119568 (16/11/2026): every Junos OS Evolved version is affected by CVE-2026-98432, the Linux kernel flaw where a child socket frees IP options inherited from a request socket with a plain kfree() while another CPU may still read them (CVSS 3.1 9.8 NVD and CNA, Red Hat rates it Moderate; EPSS under 1%; not in KEV). CERT-FR lists remote code execution and denial of service as risks. Juniper says a fix "will be available soon" and gives a workaround meanwhile (the bulletin page did not load here). Linux stable fixes exist (5.10.272, 5.15.223, 6.1.190, 6.6.159, 6.12.113, 6.18.56, 7.2.10). No exploitation mentioned. CERT-FR (source) · OpenCVE (source)
🟡 Worth knowing

GitLab 19.5.6, 19.4.8 and 19.3.9 fix 37 flaws

Added after publishingGitLab37 flawsCVE-2026-105739CVE-2026-18241

WhatGitLab fixed 37 flaws, including stored XSS, protected CI/CD variable access through a race and cross-namespace Kubernetes Secret reads.

DoUpgrade self-managed GitLab to 19.5.6, 19.4.8 or 19.3.9 now.

Details and sources
GitLab's patch release of 16 November (relayed by CERT-FR CERTFR-2026-AVI-1552) fixes four High-severity issues: stored XSS in wiki pages CVE-2026-105739 and in Error Tracking CVE-2026-108153 (both CVSS 8.7, an authenticated user runs JavaScript in viewers' browsers), access to protected CI/CD variables through a race condition in pipeline jobs CVE-2026-18241 (8.2), and reading Kubernetes Secrets from other namespaces through workspaces CVE-2026-97187 (7.6). The other 33 are Medium and Low authorization, XSS, SSRF and denial-of-service fixes, many in Duo features: CVE-2026-115428, CVE-2026-85778, CVE-2026-112203, CVE-2026-11768, CVE-2026-16609, CVE-2026-118961, CVE-2026-121533, CVE-2026-16545, CVE-2026-18491, CVE-2026-115736, CVE-2026-115218, CVE-2026-104902, CVE-2026-104934, CVE-2026-108150, CVE-2026-108155, CVE-2026-104717, CVE-2026-102907, CVE-2026-125920, CVE-2026-108154, CVE-2026-129253, CVE-2026-129321, CVE-2026-123142, CVE-2026-122045, CVE-2026-128849, CVE-2026-115219, CVE-2026-129252, CVE-2026-111747, CVE-2026-126141, CVE-2026-101358, CVE-2026-121948, CVE-2026-110990, CVE-2026-122226, CVE-2026-110750. GitLab.com is already patched and Dedicated needs no action; self-managed instances should upgrade now. No exploitation mentioned. CERT-FR (source) · GitLab (source)
🔴 Exploited or active

TYPO3 fixes 10 flaws across five supported branches

Added after publishingTYPO3CVE-2026-125160exploitedCVE-2026-119623

WhatTYPO3 fixed ten flaws, one already exploited: a Form Framework error reveals the upload path, enabling XSS through uploaded files.

DoUpdate to TYPO3 10.4.63, 11.5.57, 12.4.52, 13.4.40 or 14.3.12 and serve the upload folder with a restrictive CSP.

Details and sources
CERT-FR advisory CERTFR-2026-AVI-1553 (17/11/2026) relays ten TYPO3 GitHub advisories of the same day (CVE-2026-119621, CVE-2026-119622, CVE-2026-119623, CVE-2026-125159, CVE-2026-125160, CVE-2026-125161, CVE-2026-125162, CVE-2026-125163, CVE-2026-125164, CVE-2026-125165) with risks ranging from data exposure, security bypass, SSRF and XSS to denial of service and remote code execution; the highest is CVE-2026-119623 in the install tool (CVSS 4.0 9.4). TYPO3 says one of them is already exploited: CVE-2026-125160 (TYPO3-CORE-SA-2026-029, CVSS 4.0 6.3, CWE-209) makes the Form Framework show the raw error when saving an uploaded file fails, revealing the upload folder's path so an attacker can find the public address of files they upload, and XSS follows if the form accepts HTML or other browser-run files and that folder has no restrictive Content-Security-Policy. The advisory and the CVE record state "This vulnerability is known to be actively exploited" with no date or indicators; CISA's SSVC still says none and it is not in KEV. Fixed in 10.4.63, 11.5.57, 12.4.52, 13.4.40 and 14.3.12; also serve the upload folder with a restrictive CSP (the shipped .htaccess does it on Apache, Nginx needs manual setup). CERT-FR (source) · TYPO3 (source)
🟡 Worth knowing

OpenSSL 4.1.1, 4.0.4 and 3.5.10 fix three Low-severity bugs

Added after publishingOpenSSLCVE-2026-35190CVE-2026-107493CVE-2026-124709

WhatOpenSSL fixed three low-severity bugs: a TLS 1.3 out-of-bounds read, a PEM use-after-free and QUIC server memory growth.

DoUpdate to OpenSSL 4.1.1, 4.0.4 or 3.5.10.

Details and sources
an out-of-bounds read in TLS 1.3 key-share handling when a server uses the Suite B certificate flag CVE-2026-35190 (can crash hardened builds; 3.0, 1.1.1 and 1.0.2 not affected), a use-after-free in PEM_X509_INFO_read_bio_ex() when appending to a caller-supplied stack fails CVE-2026-107493 (only apps that pass a non-NULL stack; also fixed for premium customers in 3.0.24, 1.1.1zk, 1.0.2zt), and unbounded memory growth in a QUIC server when a client keeps opening and resetting streams CVE-2026-124709 (denial of service). OpenSSL advisory of 17/11/2026, relayed by CERT-FR CERTFR-2026-AVI-1554. No exploitation mentioned. CERT-FR (source) · OpenSSL (source)
🟠 High impact

Microsoft SharePoint: out-of-band fixes for a deserialization RCE and a privilege escalation

Added after publishingCVE-2026-124223CVSS 8.8SharePointout of band

WhatMicrosoft fixed an authenticated deserialization RCE in SharePoint Server Subscription Edition and a privilege escalation in SharePoint and Microsoft 365 Apps.

DoPatch internet-facing SharePoint first, to Subscription Edition 16.0.20326.20278 or later.

Details and sources
published 16 November, outside Patch Tuesday. CVE-2026-124223 (CVSS 3.1 8.8, CWE-502) lets an authenticated attacker run code over the network on SharePoint Server Subscription Edition before 16.0.20326.20278; CVE-2026-124222 (6.5, CWE-285, CISA-ADP technical impact total) lets an authenticated attacker elevate privileges in SharePoint and Microsoft 365 Apps for Enterprise. Microsoft's vectors mark exploit code as unproven (E:U); neither is in KEV, and CERT-FR (CERTFR-2026-AVI-1555, 17/11/2026) mentions no exploitation. Patch internet-facing SharePoint first. CERT-FR (source) · OpenCVE (source)

Breaches / Incidents 4

🔴 Exploited or active

Hijacked Optimizely scripts served a card skimmer, a crypto clipper and iOS exploit kits from 178 major sites

Optimizelysupply chainCorunaDarkSword

WhatAn attacker injected a loader into Optimizely-hosted snippets of 178 organisations, serving a card skimmer, crypto clipper and iOS exploit kits on Nov 16.

DoHunt for 1688360[.]cc, webshell[.]lol, 5vobauua3ojpklx[.]icu, eil786npto9u5bq[.]icu and 136.113.106[.]136.

Earlier coverage: 06/1105/1104/11
Details and sources
on Monday 16 November an attacker who, Optimizely says, "exploited a vulnerability in Experimentation" wrote a loader into the "Project JavaScript" of customer snippets served from Optimizely's own CDN. Report URI checked 1,611 snippets and found 188 infected, belonging to 178 organisations including CNN, the BBC, Zoom, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger and Peacock, plus Optimizely's own site and login page (the true count is likely higher). Timeline (UTC): first injection in the BBC snippet at 04:21, card skimmer (number, expiry, CVV from the site's own forms; Stripe and Adyen iframes not readable) at 13:43, an iframe loading the Coruna (iOS 13 to 17.2.1) and DarkSword (iOS 18.4 to 18.7) exploit kits from 17:13, mass cleanup from about 20:40, and only TalkTalk still infected at 21:52. The payload also swapped crypto wallet addresses and sent cookies and screenshots out; wallets took about $7,040. Cynet says the iOS delivery was conditional and not a confirmed iPhone compromise. Optimizely says the flaw is fixed and accounts secured, but has not said how it worked or how many projects were changed. No CVE, no attribution ("OPTIMJACK" is a researcher's label); the same injector was seen earlier through Kissmetrics, StorifyMe and Minute Media scripts. IOCs: 1688360[.]cc, webshell[.]lol, 5vobauua3ojpklx[.]icu, eil786npto9u5bq[.]icu, 136.113.106[.]136. The Hacker News (source) · Report URI (source)
🟠 High impact

Infoblox confirms hackers reached company data through a third-party vendor

Infobloxthird-party vendorSharePoint migration

WhatInfoblox confirmed hackers reached company information through a third-party vendor, with no indication its products or build processes were involved.

Details and sources
the networking and DNS security company says it is investigating "unauthorized access to certain company information" and has "no indication" that its products, cloud services, software updates or build-and-release processes were involved. Two people told TechCrunch the vendor does SharePoint data migration and that Infoblox has ended its work with it; one said sensitive customer data was taken. Infoblox would not say how many customers are affected or whether the attackers made demands; no group has claimed it. TechCrunch (source)
⚪ Context

US bank outages on Monday were a technical fault, not an attack

US banksoutagenot cyber

WhatLogin, transfer and ATM outages at several large US banks were a technical fault, not an attack.

Details and sources
Chase, Bank of America, U.S. Bank, Wells Fargo, Capital One and Truist customers had login, balance, transfer and ATM problems (nearly 4,800 Downdetector reports for Chase alone by 2 p.m. ET). Chase called it a technical bug, "not cyber"; Zelle said its network ran normally, and no link to last week's Microsoft 365 outage was established. Security Affairs (source)
🟡 Worth knowing

Optimizely follow-up: all monitored files clean by Monday night, and what visitors should do

Added after publishingOptimizelyfollow-upvisitor advice

WhatAll monitored Optimizely files were clean by late Nov 16, but the root cause, CVE and attribution are still unknown.

DoVisitors of affected sites on Nov 16 should update iOS, verify crypto addresses used that day and check card activity.

Details and sources
Malwarebytes adds that Report URI saw the BBC snippet clean at 02:21 UTC on 16 November, injected at 04:21 and carrying the main payload by 08:28, which does not prove the BBC's own servers were breached, and that every monitored file was clean by 23:11 UTC and stayed clean. About 190,000 live sites use Optimizely (BuiltWith), which is not the number affected. Optimizely confirmed it fixed the flaw and removed the code; still no root cause, CVE or attribution. Advice for visitors of affected sites on 16 November: update iOS or iPadOS, verify any crypto address used that day with the recipient, check card activity (Stripe or Adyen hosted forms were out of reach), and watch for odd account activity since some cookies could be stolen. Site owners should monitor what third-party scripts do and restrict what they can load. Malwarebytes (source)

Malware / Ransomware 3

🟠 High impact

BlackNevas wipes a VMware estate using only stolen credentials and admin tools

BlackNevasMeshAgentVMware vCenterESXi

WhatBlackNevas operators used stolen credentials and admin tools to take over vCenter and backups, then encrypted and wiped a VMware estate.

DoHunt for api.cloudvflare[.]com, 45.149.154[.]135, 91.92.242[.]121 and a fake vmware-vdma systemd service.

Details and sources
Kaspersky (Securelist report of 13 November) describes a 10-day intrusion in which BlackNevas operators used credentials stolen outside the virtualisation layer to reach vCenter, took over the backup servers and pulled their stored credentials, logged in to the disaster-recovery vCenter as root over SSH, deployed MeshAgent 43 seconds later and then a Ligolo-ng tunnel behind a fake "vmware-vdma" systemd service. They used vCenter Host Profiles to enable SSH on the ESXi hosts, delivered the payload over SFTP, disabled antivirus, stopped backups and encrypted the recovery systems first; all VMs on the DR host stopped within 23 seconds and roughly the first 2 KB of each virtual disk was overwritten. No CVE was exploited. IOCs: api.cloudvflare[.]com, 45.149.154[.]135, 91.92.242[.]121. Cyber Security News (source) · Securelist (source)
🟠 High impact

PerfectPioneer and KUKURAT: HTTP/3-over-QUIC implants against Southeast Asian governments

TGR-STA-0049REF7707PerfectPioneerKUKURAT

WhatA cluster linked to Squidoor attacks Southeast Asian governments through Exchange and web servers, deploying RATs that talk HTTP/3 over QUIC.

Details and sources
Unit 42 tracks TGR-STA-0049 (also REF7707, Earth Alux, Ink Dragon, linked to Squidoor/FINALDRAFT) attacking governments in Southeast Asia since at least July 2025, after earlier work in South America and Europe. Entry is through unpatched Exchange and web servers (ASPXSpy, Godzilla, Suo5 web shells) or phishing with lookalike government domains and fake PDFs, one impersonating Myanmar's election commission. PerfectPioneer is a Go RAT (six variants) with keylogging, screenshots, SOCKS5 and an "isDangerousCommand" blocklist that stops operators running noisy commands; KUKURAT is a C++ backdoor that loads at least 12 modules (BOF execution, SMB brute force, PsExec lateral movement) and a "camouflage" engine that generates decoy activity. Both talk HTTP/3 over QUIC (UDP), falling back to HTTP/2. The group dumps LSASS via comsvcs.dll by ordinal, runs DCSync and Mimikatz, spreads through logon scripts and grants itself FullAccess to every Exchange mailbox. No CVE named. Unit 42 (source)
🟠 High impact

GitDoor and GraphDoor hide their command channel in GitHub and SharePoint (Operation ShadowCloud)

Added after publishingGitDoorGraphDoorOperation ShadowCloud

WhatA campaign against Asian and African government bodies uses backdoors that hide C2 in GitHub repositories and SharePoint via the Graph API.

DoHunt for 27.102.66[.]162 and the GitHub repositories LatonyCharnleyaW/Havo and LatonyCharnleyaW/Trix.

Earlier coverage: 11/1110/11
Details and sources
Kaspersky tracks a campaign, found in early 2026, against government bodies in Asia and Africa, with an unnamed Taiwanese organisation as the first target. An abused MeshCentral agent, possibly left over from an earlier infection, fetches a legitimate binary, a sideloaded loader DLL (jli.dll, SbieDll.dll, python39.dll, WINMM.dll or libcef.dll; RC4 plus XOR, or XOR alone) and an encrypted payload that is injected in memory. GitDoor reads its commands from one GitHub repository and posts results to another (26 commands: files, keylogging, proxying, shellcode, Mimikatz and CookieKatz credential theft, WhatsApp key extraction, NTLM coercion, account creation); GraphDoor talks to SharePoint through the Microsoft Graph API with a malicious app, with plugins for keylogging, screenshots and SOCKS5 tunnelling over Ably. The GitHub repositories date from late 2024. No CVE. The actor is unidentified: Kaspersky sees overlap with Sophos's Bronze Sandcastle and similarities with FamousSparrow and CopperTurtle (BlackTech), with low confidence. IOCs: 27.102.66[.]162 (WebSocket on 443, path agent.ashx), GitHub LatonyCharnleyaW/Havo and LatonyCharnleyaW/Trix. The Hacker News (source)

Threat Actor Activity 2

🟠 High impact

Kimsuky poses as a lawmaker's aide and plants NodeLinkRAT through a note-taking service

KimsukyNodeLinkRATSouth Korea

WhatKimsuky posed as a lawmaker's aide and used an LNK lure and a note-taking service to deliver NodeLinkRAT to a seminar attendee.

Details and sources
Genians assesses with medium confidence that Kimsuky targeted someone who attended a South Korean National Assembly seminar: from 12 to 18 September the attacker asked for honorarium documents and, when told the first email had no attachment, sent an archive with an LNK disguised as an HWP form. Hidden PowerShell installs a scheduled task that checks an attacker page on an online note-taking service every 38 minutes and runs whatever code is posted there: a RAT in June, stealer and document-exfiltration code in September, and the NodeLinkRAT installer in October (it silently installs Node.js, runs commands, steals or deletes files and views the screen live). PC data went to a cloud service and documents to a free Korean web host in near real time. No CVE. Seoul Economic Daily (source)
🟡 Worth knowing

Google GTIG: agentic attacks have crossed "a significant threshold"

Google GTIGARTEXCAIRNagentic attacks

WhatGoogle says financial and suspected China-aligned actors now run multi-stage intrusions with autonomous AI agent frameworks and resell stolen LLM credentials.

Earlier coverage: 16/1113/1112/11
Details and sources
Google Threat Intelligence Group says financially motivated actors and suspected China-aligned espionage groups now run multi-stage intrusions with autonomous multi-agent frameworks (ARTEX, CAIRN, CyberStrikeAI, PentAGI, Agent Zero), use the ZeroHollow inference service for exploit development, and diff patches with LLMs to weaponise fixes quickly (its example: the September PaperCut NG/MF flaws, no CVE ID given). Stolen LLM credentials are resold through "LLM mixer" proxies such as APIMart and JetNight, and the Chinese-language Jieyou phishing-as-a-service kit adds an LLM chatbot to customise phishing panels. No named APT or CVE. The Hacker News (source) · Google (source)

Law Enforcement 3

🟡 Worth knowing

Alleged 8-Digits Team ransomware member extradited from Italy to the US

8-Digits TeamINC RansomLynxextradition

WhatAn alleged 8-Digits Team ransomware member was extradited from Italy to the US, charged over attacks on schools and healthcare.

Details and sources
Roman Khlynovskiy, 43, a Kazakh national, appeared in federal court in Knoxville, Tennessee, charged with conspiracy to commit computer fraud and identity theft, conspiracy to commit wire fraud, extortion involving computers, conspiracy to commit money laundering and threats to disclose nonconsensual intimate images under the TAKE IT DOWN Act (up to 27 years). Prosecutors say the 8-Digits Team, affiliated with INC Ransom and Lynx, hit schools, hospitals and healthcare entities in at least six states, claimed over 50 TB of data including more than 70 million health records, and demanded more than $500 million in ransoms. The FBI investigated, with help from Ukraine's State Bureau of Investigation and Prosecutor General's Office and from Lithuania; the arrest date in Italy was not given. These are allegations. DOJ (source) · BleepingComputer (source)
🟡 Worth knowing

Lapsus$ hacker Arion Kurtaj pleads guilty, sentencing on 23 November

Lapsus$Arion Kurtajguilty plea

WhatLapsus$ member Arion Kurtaj pleaded guilty to computer misuse and fraud charges, with sentencing set for Nov 23.

Details and sources
at Woolwich Crown Court the Lapsus$ member, now 21 and 16 to 17 at the time of the offences, admitted three counts of unauthorised acts impairing a computer, one of unauthorised access with intent to commit further offences, one of fraud and possessing a mobile phone in prison, plus assaults on custody staff; no blackmail conviction was listed. The hacks include Rockstar Games (the 2022 leak of 90 GTA VI clips, done from a hotel while on bail), Uber (over $2.8 million in costs), Revolut, BT and EE. Gizmodo (source) · IBTimes UK (source)
⚪ Context

Italy fines Meta €805 million over Facebook's camera roll suggestions

Added after publishingItalyGaranteMeta€805 million

WhatItaly fined Meta €805 million because Facebook's camera roll suggestions read the photo galleries of about 20 million Android users without clear notice.

DoCheck Facebook's photo access under Android Settings > Apps > Facebook > Permissions, or iPhone Settings > Privacy & Security > Photos.

Details and sources
Italy's data protection authority, the Garante, says Facebook's "camera roll suggestions" feature (launched in 2023, on by default) read the whole photo gallery of about 20 million Android users in Italy under a single app-wide permission whose notice did not make the feature clear, breaching Article 122 of the Italian Data Protection Code (the ePrivacy rules). The probe began in July 2025 after a complaint by the consumer group CTCU; the Garante refused Meta's request to hand the case to Ireland's regulator, says the violation is still ongoing for some users and cites Meta's "poor level of cooperation". Meta has 60 days to ask consent separately for the feature, explain its use and switch it off by default, including for users who gave blanket permission; the Garante left open whether camera roll content is used to train AI, which Meta denies. Meta calls the fine disproportionate and will appeal. To check your phone: Android Settings > Apps > Facebook > Permissions > Photos and videos; iPhone Settings > Privacy & Security > Photos. Malwarebytes (source)
‹ 16/11/202618/11/2026 ›