Koffei

Daily CyberNews · 28/10/2026

‹ 27/10/202629/10/2026 ›
Today in 30 seconds
  • 🔴 Update older iPhones and iPads to iOS/iPadOS 18.7.11: CVE-2026-86950 is exploited.
  • 🔴 Patch Brotli copies everywhere: CVE-2026-111497 is now in KEV with exploit attempts.
  • 🔴 Patch Magento/Adobe Commerce: PolyShell CVE-2026-48356 is in KEV, 953 stores infected.
  • 🔴 Patch PeopleSoft for CVE-2026-35273: ShinyHunters chain exploited again in September.
  • 🔴 Block Microsoft 365 device-code flow where unused: Skyw4lker phishing is active.

Vulnerabilities 10

🔴 Exploited or active

Apple backports CoreGraphics zero-day fix to older iPhones/iPads

CVE-2026-86950CVSS 8.8Apple CoreGraphics

WhatApple backported its fix for an exploited CoreGraphics out-of-bounds write, used in targeted attacks, to older iPhones and iPads.

DoUpdate older iPhones and iPads to iOS/iPadOS 18.7.11, and newer devices to 26.7.1 or later.

Details and sources
Apple backported its fix for CVE-2026-86950 (CVSS 8.8, an out-of-bounds write in CoreGraphics letting arbitrary code run when a crafted file is processed, reported by Meta's Product Security team, originally fixed in iOS/iPadOS/macOS 26.7.1 in September) to iOS/iPadOS 18.7.11 for older devices (iPhone XS/XS Max/XR, older iPads) on Oct 26, citing a report of exploitation "in an extremely sophisticated attack against specific targeted individuals" on pre-iOS 27 versions. The Hacker News (source) · Help Net Security (source)
🟠 High impact

GoAhead embedded web server TLS flaws hit hundreds of millions of IoT, firewall, OT and military devices

Added after publishingGoAheadCVE-2026-115881CVE-2026-115884IoT/OT

WhatThree GoAhead TLS flaws let revoked client certificates pass mTLS, bypass chain-depth checks or make DHE sessions decryptable; no exploitation confirmed.

DoAsk device vendors for firmware with GoAhead 6.1.1 and prefer ECDHE or TLS 1.3 over DHE_RSA suites.

Details and sources
researcher Sébastien Féry (FoxIO) detailed three flaws in the widely-embedded GoAhead web server. CVE-2026-115881 (GOAHEAD-SA-2026-0006, CVSS 9.1) — OpenSSL-backend builds load a CRL but never set the revocation-check flags, so a revoked CA-signed client certificate still completes an mTLS handshake. CVE-2026-115884 (GOAHEAD-SA-2026-0004, CVSS 7.4) — a LibreSSL-backend legacy certificate-verifier bug lets a 17-certificate chain pass (tested against LibreSSL 3.2.x/3.3.2–3.3.6), effectively bypassing mTLS chain-depth checks. A third, no-CVE flaw (GOAHEAD-SA-2026-0005, CVSS 5.9, Mbed TLS builds using TLS 1.2 DHE_RSA suites) truncates the 2048-bit Diffie-Hellman modulus to a single byte, leaving only 8 possible shared secrets — a passive observer who sees DHE-negotiated TLS traffic can brute-force the session key and decrypt logins/session tokens; a public proof-of-concept exists, and FoxIO reproduced key recovery in a lab in 5 attempts (ECDHE and TLS 1.3 are unaffected). FoxIO names Palo Alto Networks, Comcast, D-Link, HP, Canon, Honeywell, GE, Oracle, Siemens, Rockwell, Digi, Lockheed Martin and Boeing among the organizations whose products use GoAhead, though it doesn't confirm which ship an affected build/configuration. Vendors were notified in April 2026; GoAhead fixed all three in 6.1.1 (given to device makers Aug 2, publicly published Oct 12), but device owners only get the fix once their manufacturer ships a firmware update, and it's unclear whether any have yet. No confirmed in-the-wild exploitation. FoxIO (source) · The Hacker News (source)
🟡 Worth knowing

Microsoft Authenticator for Android let rogue apps steal sign-in tokens

CVE-2026-80097CVSS 8.6Microsoft Authenticator

WhatA signature-check bypass let a permissionless Android app impersonate Teams and steal work-account tokens from Authenticator 6.0.0 onward.

DoUpdate Microsoft Authenticator for Android to 6.2606.4246.

Details and sources
CVE-2026-80097 (CVSS 8.6, AV:L/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H) let a permissionless Android app impersonate a trusted app, such as Teams, and steal work-account sign-in tokens because of a signature-check bypass (affecting Authenticator versions 6.0.0 up to but not including 6.2606.4246, per Microsoft's own advisory); fixed in 6.2606.4246. Microsoft rates exploitation "less likely" and says the flaw was not publicly disclosed before the patch; EPSS score is under 1%. The Hacker News (source)
🟠 High impact

Critical Wasmtime/Cranelift sandbox-escape flaws

CVE-2026-115889CVE-2026-115893CVSS 9.9Wasmtime

WhatTwo Cranelift flaws could let malicious WebAssembly escape the Wasmtime sandbox; both are fixed and not exploited.

DoUpdate Wasmtime to 36.0.19 or the fixed 48.0.7 or 49.0.4 releases.

Details and sources
CVE-2026-115889 (9.9, all architectures, affecting Wasmtime before 36.0.19 and 37.0.0 up to 43.0.0) and CVE-2026-115893 (9.9, ARM64 builds using the default Cranelift backend, before 36.0.19 and in ranges up to 48.0.7/49.0.4) could let malicious WebAssembly code escape the Wasmtime sandbox. Both are fixed upstream, neither is in CISA's KEV catalog, and no active exploitation has been reported; EPSS scores are under 1% for both. The Hacker News (source)
🔴 Exploited or active

Brotli decompression flaw ("Node2Shell"/"Broccoli2Shell") hits Node.js, .NET, Bun, Envoy, Istio and Python — now in CISA's KEV catalog

Added after publishingCVE-2026-111497CVSS 9.2in KEVBrotli

WhatThe Brotli decoder flaw in bundled copies across Node.js, PHP, Bun, Envoy, Istio, .NET and Python was added to KEV, with exploit attempts logged.

DoPatch each bundled copy: Node.js 22.23.4, 24.22.1 or 26.12.1, Bun 1.4.4, PHP 8.x-1 builds; federal deadline Nov 10.

Details and sources
CVE-2026-111497 (CVSS 9.2), an unchecked-return-value flaw in Google's Brotli decoder enabling memory corruption, data disclosure, denial of service or possibly code execution from a crafted stream, affects any runtime or proxy that bundles its own copy of the library — Node.js (fixed 22.23.4/24.22.1/26.12.1), PHP (fixed 8.5.12-1/8.4.27-1/8.3.36-1/8.2.35-1), Bun (fixed 1.4.4), Envoy Proxy, Istio, .NET, ClickHouse and Python's brotli/brotlicffi packages, among others — since each ships its own vulnerable copy and needs separate patching. CISA added it to the Known Exploited Vulnerabilities catalog on Oct 27 alongside the Adobe Commerce PolyShell flaw (see below), giving federal agencies until Nov 10 to patch; honeypot operators Previdian and ELLIO have already logged exploit attempts. Wiz (source) · The Hacker News (source)
🟡 Worth knowing

Tailscale patches "Tailworm" SSH flaws in version 1.104.4

Added after publishingTailscaleTailworm1.104.4

WhatTailscale 1.104.4 fixes SSH socket-forwarding flaws that let a permitted session reach root-only sockets; Headscale 0.29.5 is still unpatched.

DoUpgrade Tailscale to 1.104.4, or set TS_SSH_DISABLE_FORWARDING=1 on unpatched nodes.

Details and sources
Tailscale shipped 1.104.4, covering four bulletins (TS-2026-014 through TS-2026-017) for the SSH socket-forwarding flaws researcher Erica Windisch (Cyberdione Labs) disclosed Oct 25 — the worst let a permitted SSH session reach root-only sockets, such as the Docker socket, via a timing-based path swap. No CVE has been assigned. Windisch called the chain "wormable" ("Tailworm"); Tailscale disputes that framing, saying the findings involve lateral movement within policy-allowed connections rather than self-replicating malware. Headscale, the self-hosted control server, remains unpatched as of its latest release (0.29.5). Interim mitigation for unpatched nodes: set TS_SSH_DISABLE_FORWARDING=1. No in-the-wild exploitation reported. The Hacker News (source)
🔴 Exploited or active

ShinyHunters' exploited Oracle PeopleSoft zero-day is three chained flaws

Added after publishingShinyHuntersCVE-2026-35273in KEVPeopleSoft

WhatThe PeopleSoft flaw ShinyHunters exploited is a three-bug chain from PSIGW SSRF to code execution, hit again in a September wave.

DoApply Oracle's June 10 out-of-band PeopleSoft fix for PeopleTools 8.61 and 8.62.

Details and sources
Mandiant (tracking the group as UNC6240) and the Zero Day Initiative found that the critical Oracle PeopleSoft issue ShinyHunters exploited is actually a chain: an unchecked SSRF in the Integration Gateway (PSIGW) that makes attacker requests look like they originate from localhost, which then passes the Environment Management Hub's (PSEMHUB) trust check, leading to code execution via a web shell or a malicious XML file loaded through Java's XMLDecoder. Only one CVE exists so far — CVE-2026-35273 (CVSS 9.8, unauthenticated network takeover, affecting PeopleTools 8.61/8.62) — which is in CISA's KEV catalog; ZDI is working with Oracle to assign separate IDs to the other two flaws. An initial wave (May 27–June 9) hit mostly universities as a true zero-day; a September follow-up hit tech, IT services, healthcare, agriculture, transport and government targets worldwide, bypassing WAF rules with path-encoding tricks. Oracle shipped an out-of-band fix on June 10; ShinyHunters is known to steal data and extort victims. The Hacker News (source)
🟡 Worth knowing

Anthropic says an unreleased internal research model chained flaws into working exploits for Chrome, Safari, Windows and Linux

Added after publishingAnthropicAI-found exploitsCVE-2026-98432

WhatAnthropic says an unreleased model found and chained flaws into exploits for Chrome, Safari, Android, Windows (CVE-2026-33096) and Linux, all now patched.

Details and sources
Anthropic's Frontier Red Team reported that an unreleased model discovered and chained vulnerabilities across Chrome (ending in a sandbox-escape via a sign-in use-after-free), Safari (disabling same-origin policy), Android (data extraction from a locked, stolen phone), Windows (CVE-2026-33096, a network-triggered denial-of-service, patched April 2026) and Linux (CVE-2026-98432, an unauthenticated root shell over a multipath-TCP listener, patched Oct 15) — all built and reported privately, with no real-world attack use, and all now patched. Anthropic reported 62 flaws to Google alone; Apple and Google credit Claude/Anthropic researchers on some of the resulting fixes, though the article notes some of Anthropic's CVE linkages (particularly for the Windows and Linux chains) don't precisely match the public advisory text. The Hacker News (source)
🟠 High impact

Adobe fixes 8 critical Campaign Classic flaws, plus a separate critical Connect bug

Added after publishingAdobe Campaign ClassicCVE-2026-104351CVE-2026-48360

WhatAdobe fixed eight critical Campaign Classic flaws up to CVSS 9.9 and a critical Connect code-injection bug; no exploitation known.

DoUpgrade Campaign Classic to 7.4.4 build 9403 and Connect Desktop to 2026.9.61, 2026.10.66 or 2026.10.170, Connect Mobile Android to 4.5.

Details and sources
Adobe patched eight critical vulnerabilities in Campaign Classic (versions 7.4.4 build 9402 and earlier, on Windows and Linux; fixed in build 9403) — CVE-2026-104351 and CVE-2026-104353 (both CVSS 9.9, code/OS command injection enabling code execution), CVE-2026-105780 (9.9, incorrect authorization enabling code execution), CVE-2026-105489 (9.9) and CVE-2026-105781 (9.8, SQL injection enabling security-feature bypass), CVE-2026-104350 (9.3, reflected XSS enabling privilege escalation), and CVE-2026-104354/CVE-2026-105779 (9.1 each, SQL injection). Adobe also separately fixed CVE-2026-48360 (CVSS 9.6) in Connect Desktop/Mobile, a critical code-injection flaw allowing arbitrary code execution (fixed in Connect Desktop 2026.9.61/2026.10.66/2026.10.170 and Connect Mobile for Android 4.5). Adobe says it isn't aware of any in-the-wild exploitation of either set. The Hacker News (source)
🟠 High impact

Chrome and Firefox ship large security updates — install them

Added after publishingChrome 156.0.8078.73Firefox 159.0MFSA2026-118

WhatChrome fixes 52 security bugs, 4 critical, and Firefox 159.0 fixes 167, including use-after-free bugs that can allow code execution.

DoUpdate Chrome to 156.0.8078.73/.74 and Firefox to 159.0.

Details and sources
Google's Chrome 156.0.8078.73/.74 (Windows/Mac) and .73 (Linux), released Oct 27, carries 52 security fixes, 4 rated Critical; Mozilla's Firefox 159.0, covered by advisory MFSA2026-118, fixes 167 vulnerabilities. Both releases patch use-after-free bugs in page layout, JavaScript processing and SVG graphics that can cause crashes or, in some cases, let an attacker run code. Google and Mozilla both credit OpenAI's Codex Security team with finding some of the bugs; Mozilla separately describes a collaboration in which Anthropic researchers validated vulnerabilities before reporting them so Mozilla could reproduce and fix them (echoing Anthropic's own disclosure above). Neither vendor's release notes given here include CVE IDs or confirm in-the-wild exploitation. Malwarebytes (source)

Breaches / Incidents 4

🟠 High impact

University of Copenhagen data breach exposes 525,000 people

University of Copenhagen525,000 people

WhatA breach of a university test environment exposed CPR numbers and contacts for about 525,000 people plus PINs for about 20,000 more.

Details and sources
a breach of a test environment at the University of Copenhagen exposed names, Danish national ID (CPR) numbers, phone numbers and emails for roughly 525,000 people, plus PIN codes for about 20,000 more. The incident is contained, under investigation, and reported to the Danish Data Protection Agency; no attacker has been identified. BleepingComputer (source)
🟡 Worth knowing

DreamHost forces password resets amid ongoing, unconfirmed incident

DreamHostforced password resetsongoing

WhatDreamHost keeps force-resetting email, FTP and SFTP passwords in an unconfirmed ongoing incident that may involve email accounts.

DoDreamHost users should check mail forwarding rules.

Details and sources
web host and domain registrar DreamHost has been force-resetting email, FTP and SFTP account passwords since Oct 23 in what it calls an "ongoing incident," while declining to confirm a breach or respond to press queries; language in its own customer notices (warning about forwarding-rule abuse) suggests possible email-account exposure. Status remains unresolved. BleepingComputer (source)
🟡 Worth knowing

Keystone to destroy and reissue 18,835 hardware wallets after website breach

Keystonehardware walletsfollow-up

WhatKeystone's website breach exposed its device-verification key and about 26,000 customer records; 18,835 unsold wallets will be reissued.

Details and sources
hardware crypto-wallet maker Keystone had its website breached via a vulnerability in a third-party monitoring tool, exposing its device-authenticity verification key and roughly 26,000 customer records (names, shipping/contact details, orders — no payment or ID data). Keystone detected and blocked the intrusion on Oct 20, fixed the vulnerability, and is destroying and reissuing 18,835 unsold wallets with a new key; no counterfeit devices or thefts have been confirmed. The Hacker News (source)
🟠 High impact

Japan's GMO offers $1 million bounty after refusing $50 million extortion demand

Added after publishingGMOOnamae Desktop Cloudextortionbounty

WhatGMO refused a $50M extortion demand over its Desktop Cloud breach, saw leaked data and unauthorized OANDA trades, and offers a $1M bounty.

Details and sources
GMO Internet Group discovered Oct 20 that attackers had planted malware on customer servers in its Onamae.com Desktop Cloud (FX-trading) service, affecting up to 26,739 servers; GMO shut the service down Oct 21 and refused a roughly $50 million (¥8 billion) extortion demand. Broker OANDA Japan confirmed unauthorized trades using credentials believed leaked from GMO's servers, and attackers leaked what they called a customer's personal data on Oct 27. GMO launched a "Cybercrime Bounty" on Oct 28 — up to $1 million, payable only on a qualifying tip leading to an arrest or an international wanted notice — and is compensating affected customers directly. The Hacker News (source)

Malware / Ransomware 4

🟠 High impact

"Aokigahara"/"Aoki" infostealer adds a module targeting AI coding assistants

AokigaharainfostealerAI sessionsClaude

WhatThe Aoki Windows infostealer added a module stealing local sessions for Claude, ChatGPT Desktop, Codex and Cursor plus Git, SSH and cloud credentials.

Details and sources
a Windows infostealer gained a dedicated aisessions module that collects local authentication/session data for Claude, ChatGPT Desktop, Codex and Cursor, alongside Git/SSH/cloud (AWS/Azure/GCP) credentials, browser data and crypto wallets; the AI-targeting capability appears only in newer samples. Check Point Research published the analysis Oct 23 (Symantec/Carbon Black first flagged the stealer Oct 15); no confirmed breach of the AI providers themselves or established monetization path. Gridinsoft (source)
🟠 High impact

"uCreater" rentable Android spyware platform targets Telegram, WeChat and mobile wallets

uCreaterAndroid spywareTelegramWeChat

WhatA rentable Android spyware platform uses app virtualization to intercept OTPs and 2FA codes, read Telegram databases and spoof camera feeds.

Details and sources
a new multi-tenant Android spyware-for-rent platform uses app-virtualization tooling (VirtualApp/LSPatch/Pine) to intercept OTPs, query live Telegram databases, capture cleartext 2FA codes and spoof camera feeds; a possible link to a Changsha/Hunan-based vendor is called "plausible but unproven." Security Affairs (source) (independent researcher "Ovi," relayed Oct 26)
🔴 Exploited or active

Magento/Adobe Commerce JS-injection campaign hits 953 online stores — PolyShell now in CISA's KEV catalog

Added after publishingPolyShellCVE-2026-48356in KEVMagento

WhatA scanner-evading JavaScript loader in cms_block infected 953 Magento/Adobe Commerce stores via PolyShell and SessionReaper (CVE-2025-54236).

DoPatch Magento/Adobe Commerce by the Nov 10 deadline and inspect the cms_block table for injected scripts.

Details and sources
researchers at Sansec found 953 Magento/Adobe Commerce stores infected with a scanner-evading JavaScript loader stashed in Magento's cms_block database table, entering mostly via the PolyShell (CVE-2026-48356, CVSS 9.3, an unrestricted file-upload flaw) and SessionReaper (CVE-2025-54236) flaws; the campaign has been running since at least July 2025 and no next-stage payload has been recovered, so its ultimate goal is unclear. CISA added PolyShell to its Known Exploited Vulnerabilities catalog on Oct 27 (alongside the Brotli flaw above), giving federal agencies until Nov 10 to patch. No actor has been named. The Hacker News (source) · Security Affairs (source) · The Hacker News (source)
🔴 Exploited or active

Microsoft Teams "IT support" vishing scam delivers Chaos ransomware and a new GoGRPC backdoor

STAC4749GoGRPCChaos ransomwareTeams

WhatSTAC4749 impersonates IT support in Teams external chat to install RMM tools, then deploys the GoGRPC backdoor and Chaos ransomware.

DoRestrict Teams external chat and alert on new RMM tool installs.

Details and sources
a cluster tracked as STAC4749 impersonates IT support over Microsoft Teams external chat to get victims to install legitimate RMM tools, moves laterally through Active Directory, and deploys a new backdoor ("GoGRPC") before detonating Chaos ransomware (per Sophos). Primary research from Arctic Wolf Labs; no group name attached yet. The Hacker News (source)

Threat Actor Activity 8

🟡 Worth knowing

Lazarus Group runs "Operation FlashHole" against South Korean firms with post-quantum C2

LazarusFlashHoleSouth Korea

WhatLazarus abuses INISAFE CrossWeb EX to sideload a malware chain ending in Cubridge, with ML-KEM-1024 encrypted C2, against South Korean firms.

Details and sources
Kaspersky (GReAT researcher Sojun Ryu) reports a new Lazarus espionage campaign against South Korean construction, software and telecom-equipment firms, abusing the mandatory Korean financial security software INISAFE CrossWeb EX to sideload a malware chain (TechHub/TouchHub/TouchShift/ClientHub/Cubridge/LPEClient) — notable for using ML-KEM-1024 post-quantum encryption for command-and-control since around September 2024. It continues the Operation SyncHole/"Cascade Hole" cluster. The Hacker News (source)
🔴 Exploited or active

Dahua IP cameras hijacked en masse in Ukraine

DahuaCERT-UACVE-2021-33044CVE-2021-33045

WhatAttackers are hijacking Dahua IP cameras across Ukraine via predictable Easy4IP/P2P serials, weak credentials or old auth-bypass bugs.

DoDisable P2P on Dahua cameras, keep them off the internet and update firmware.

Details and sources
CERT-UA warns that attackers are hijacking Dahua IP cameras across Ukraine by abusing predictable Easy4IP/P2P serial numbers combined with weak credentials or older auth-bypass bugs (CVE-2021-33044/CVE-2021-33045); no attacker has been attributed, and mitigation requires disabling P2P, avoiding internet exposure, or updating firmware rather than a single patch. The Hacker News (source)
🟡 Worth knowing

Australia's ASD discloses its own disruption of the Amadey botnet backend

ASDAmadeydisruption

WhatAustralia's ASD disconnected an Amadey backend from about 180,000 infected machines and destroyed 5TB of stolen data.

Details and sources
Australia's Signals Directorate says it accessed a criminal's Amadey malware-loader backend, disconnected it from roughly 180,000 infected machines worldwide, and destroyed 5TB of stolen data before it could be sold — a separate, belated disclosure distinct from the earlier multinational "Operation Endgame" takedown in June, which did not name Australia. The Hacker News (source)
🟡 Worth knowing

Seven arrested in South Korea over AI-assisted breach-and-extortion spree

South KoreaARTEX AIseven arrested

WhatSeoul police arrested a seven-person ring using ARTEX AI that breached 90+ organizations, stole 8.2M records and extorted about $270,000.

Details and sources
Seoul police identified and arrested a seven-person ring — led by a 32-year-old former penetration tester — that used the "ARTEX AI" autonomous attack framework to breach more than 90 organizations (hospitals, large corporations, golf courses), stealing 8.2 million records and extorting roughly $270,000 across 38 incidents between arrests on Oct 15 and referral to prosecutors Oct 23. BleepingComputer (source)
🟡 Worth knowing

US arraigns extradited suspect in $36.9 million crypto investment-fraud scheme

Pig butcheringextradition$36.9M

WhatAn extradited suspect was arraigned in California on money-laundering charges tied to a Cambodia-run crypto fraud with 174 US victims.

Details and sources
Zheng Zhao, extradited from the Netherlands, was arraigned Oct 23 in the Central District of California on money-laundering conspiracy charges tied to a Cambodia-run investment-fraud ("pig-butchering") scheme with 174 identified US victims. Bitcoin.com (source)
🟠 High impact

"CyberStrikeAI" uses 44 AI agents to automate hacking against Japanese targets

Added after publishingCyberStrikeAI44 AI agentsJapan

WhatExposed servers ran CyberStrikeAI with 44 agents automating full attacks, including fake Japanese login pages, against 19 organizations; no breach is confirmed.

Details and sources
Trend Micro's TrendAI division found exposed servers running CyberStrikeAI, an open-source pentesting tool, split across 44 agents automating the full attack lifecycle — reconnaissance, exploitation, command-and-control, persistence, exfiltration and even report generation — including the ability to generate fake Japanese-language login pages. The activity, concentrated in October 2026, ran validation checks against at least 19 organizations, 16 of them in Japan; no breach is confirmed and no operator has been named. The Hacker News (source)
🟡 Worth knowing

FBI seizes Virtualine bulletproof-hosting domain over alleged CSAM hosting and bitcoin laundering

Added after publishingVirtualinebulletproof hostingFBI seizure

WhatThe FBI and European police seized the domain of bulletproof host Virtualine, alleged to host illegal material and launder Bitcoin.

Details and sources
the FBI, with Dutch, German and UK police, seized the domain of Virtualine (also operating as Secury, Massgod, Spamhouse, Lanedo and Omegatech), a "bulletproof" hosting provider the DOJ alleges knowingly hosted a forum selling child sexual abuse material and nonconsensual intimate images, and laundered proceeds through unhosted Bitcoin wallets to dodge exchange compliance checks. Chainalysis linked its infrastructure to ransomware operators and infostealer campaigns; the operator disputes knowingly hosting the illegal material. The Hacker News (source)
🔴 Exploited or active

"Skyw4lker" device-code phishing campaign hits US manufacturing, tech and healthcare firms

Added after publishingSkyw4lkerdevice-code phishingMicrosoft 365

WhatAn active campaign sends password-protected lures via sites[.]google[.]com into a real M365 device-code flow to harvest valid access tokens.

DoRestrict the Microsoft 365 device-code flow with Conditional Access and warn users about device-code prompts in documents.

Details and sources
sandbox vendor ANY.RUN detailed an active campaign, dubbed "Skyw4lker," sending victims a password-protected document (posing as a contract or service proposal) that routes them through a page on sites[.]google[.]com into a genuine Microsoft 365 device-code authentication flow — the attacker starts a legitimate device-authorization request and tricks the victim into completing it, then harvests the resulting valid Microsoft 365 access tokens, so no credential is "stolen" in the traditional sense. The phishing backend uses obfuscated JavaScript/PHP to track victim sessions and poll for completion. No specific actor or nation has been attributed. The Hacker News (source)
‹ 27/10/202629/10/2026 ›