Phish Wire

Phish Wire - Oct 7, 2026

Written by Identity Automation | Oct 20, 2026, 2:00:00 PM

The period from September 24 to October 7, 2026 produced 103 detections. Eighteen campaigns reached 37 organizations in eight states. Kentucky accounted for 53 of those detections and Florida for 24. Four campaigns captured no credentials at all. They were tech-support scareware pages with disabled inputs and a callback phone number repeated across fake dialogs. One of them produced 72 detections across 68 rotated hostnames on Azure static web hosting. Among the credential pages, three sent the password out through a Socket.IO relay instead of a form POST. Several others followed the password step with a one-time code prompt.

Two of those ran the Office360 modal-form kit, posting to processmail.php and then process.php. Seven campaigns impersonated Microsoft directly. Others used Microsoft Support, Windows Defender, Amazon, Google, AOL and invitation-service branding. Hosting ranged from rotated hostnames on b-cdn[.]net, Azure and Azure Front Door to single attacker-registered domains. Our read is that this period's pages moved away from the plain form POST. Callback scams and WebSocket relays give credential-submission rules nothing to match. Live OTP collection leaves the one-time code as exposed as the password.

Domains Reviewed

  • g3z3abhqjtvvmgw9zkwxb3r2.b-cdn[.]net/?utm_medium=paid&utm_source=fb&utm_... (5 variants)
  • h7am35794z7hgcshkkgabr4g.z13.web.core.windows[.]net/ (72 variants)
  • heemawarii-h6cdedfrfxdcgbbp.z03.azurefd[.]net/wcdbcd/index.html?bcda=1-8... (6 variants)
  • ptiry[.]click/giftx/gifrd/
  • lolafathiakunfayakunlogs.morninggoldenlight[.]de/Xl474/index.php?key=NEG...
  • btnuxyyygf.bralumekio[.]com/tiger7_computer2_flower3_floor1_butter10_che... (3 variants)
  • login.rebrand.org[.]nz/?c3cbc2=H54i6AjWn_nvGmsE7zfXPte4wdm2XF7D-g&sso_re... (2 variants)
  • hasluhprhi.jentaxilon[.]com/valley9_grass2_rice5_branch9_mountain3_thund...
  • pencil1.retroprobe[.]com/air2_tea5_flower5_city6_table8_valley8_ship1_ra...
  • anself[.]com/SB7921/secure/signin/passwd
  • login.securedistrict[.]icu/?d90cc1=iuWf2mLSdZokR06dvv-SuKGV9iA0zUoLGw&ss...
  • smtp.aybemail[.]com/
  • prime-reauth.document.doc8fs743h26ash74724h6[.]online/signin
  • 3x8rtih43c.yuansenmenye[.]com/x1eZnWmFu6J6ZpGbackK8HLaTvC5Yh/uG7WC9jhJFO... (2 variants)
  • microsoftss.silverveil[.]org/mLH7L/index.php?key=LKUGoYSkZy (2 variants)
  • info2party[.]com/check/chzaldm/chzalndm/chzadmdlz/?
  • theuniqueidea[.]com/admin/?mxid=dGFyYS5oZWRsdW5kQGRpc3RyaWN0MTk2Lm9yZw%3...
  • tytghvdrerdfc44.bluerock-8cf69a02.westus2.azurecontainerapps[.]io/wcdbcd...

A paid Facebook ad click led users at organizations in Kentucky, Texas, and Florida to a fake Microsoft Support page that simulates a locked, infected Windows session. Five detections occurred on October 2, October 5, and October 7, 2026. The page captures no credentials. The Email and Unlock Key inputs are disabled and not wrapped in a form, so nothing can be typed or submitted. The goal is a phone call. Dozens of fake dialogs such as Password required for System32 and Memory access violation at 0x88412 cover the page. The page also hides the real cursor and draws a fake one, and CSS text selection is disabled. A popup shows the visitor's IP address and urges them to confirm their identity or close the page. The campaign rotated across 5 hostnames on the b-cdn[.]net CDN platform. Because the page has no form post or exfiltration endpoint, network rules that look for credential submission will not fire. Detection has to key on the page content, the repeated phone number and the ad-click entry URL. Users are pushed to a live phone operator instead of a login form.

Related subdomain variants:

  • brightdash7606.b-cdn[.]net
  • ctph4cc6z7gb23im2d9cv8jx.b-cdn[.]net
  • g3z3abhqjtvvmgw9zkwxb3r2.b-cdn[.]net
  • kysjiskjm3zj48gf3qv6d4aj.b-cdn[.]net
  • q8u9kp2rwd3fihx3i88rd5mh.b-cdn[.]net

This campaign used a fake Microsoft Support page and a staged system-crash display as a tech-support scam aimed at organizations in Kentucky, Florida, Illinois, Georgia, Idaho, Maryland, Texas, and Minnesota. 72 detections occurred between September 24 and October 7, 2026. The page is a single-page app that fills the screen with a copy of the Microsoft Support site. On top of it sit dozens of fake dialogs reading "Memory access violation at 0x88412" and "Password required for System32". The page has no working credential capture. The Email and Unlock Key inputs are disabled, and the buttons only steer the victim toward calling the number. The script hides the cursor and shows a blue-screen effect, which pushes the user to act without testing the page. The campaign rotated across 68 hostnames on the Azure web.core.windows[.]net platform. A scareware page with no form POST gives network-based credential rules nothing to match. Detection has to rely on the hosting pattern and on the repeated callback number.

Related subdomain variants:

  • aacbypcmzcwfwt4eiahcymbd.z20.web.core.windows[.]net
  • azazcg3xwzt23m8hnfwrgdip.z20.web.core.windows[.]net
  • biaqs44qvxughxf5wsuyjgsq.z20.web.core.windows[.]net
  • ccwmndt3q8sg59b25enu8mq8.z20.web.core.windows[.]net
  • ckmntaer4si3g4n6kg3m88za.z13.web.core.windows[.]net
  • dc8mwrufwhz6xqttcm75j2jp.z20.web.core.windows[.]net
  • ddz3uq6ayy3g8cyg6dkkvnpz.z20.web.core.windows[.]net
  • dfp92tkhqf688c89rmmsvgw8.z20.web.core.windows[.]net
  • dtfa5qnbps9c9vgqqahgqd3f.z1.web.core.windows[.]net
  • e4jukarznxytyezpiumavaqg.z22.web.core.windows[.]net
  • eqagevwdvwkvg3nheaatjcfn.z20.web.core.windows[.]net
  • ffbu25kpc3vutuqcgbdtjn5e.z13.web.core.windows[.]net
  • fhte6iecdb5t48my8r4j7hcu.z20.web.core.windows[.]net
  • ftrupisb6yfm3i5xtiid9aw6.z20.web.core.windows[.]net
  • fx2qqv2zw7bngaipsnxhq7pr.z20.web.core.windows[.]net
  • g6j38r43bw34pehw5p4u8jed.z1.web.core.windows[.]net
  • gudccp8bjqcd3zjhy5sa6syr.z20.web.core.windows[.]net
  • gy3cbtmf6qkbh97akhfsisbv.z5.web.core.windows[.]net
  • h7am35794z7hgcshkkgabr4g.z13.web.core.windows[.]net
  • h8vmjmg268bkd8x9p5xxxyyq.z13.web.core.windows[.]net

This campaign used a fake Windows security lock page to push targeted organizations in Kentucky and Idaho toward a tech-support phone number. Six detections occurred between September 25 and October 7, 2026. The page does not capture credentials. The username and password inputs are read-only, and the Login, Cancel, Scan Now and Decline controls are links to # or javascript:void(0). The goal is a phone call. The same number is also passed in the URL query string. The text claims the device is locked, names a fake file called system-crash.dll and says access is blocked. It tells the user to contact an agent. A Microsoft Support Center block and a Windows Defender label lend the page a Microsoft look. The campaign rotated across 6 hostnames on the azurefd[.]net platform. Because the page has no submission target, network rules that look for credential POSTs will not fire on it. Detection has to key on the page content and the phone number. Users should treat any browser page that shows a support hotline as untrustworthy.

Related subdomain variants:

  • fgghjjiui-gkf4b8dacjbscddq.z03.azurefd[.]net
  • fuffadddabn-bkena5aubgcadbge.z03.azurefd[.]net
  • gbtryjyutft-gddud4cpbrbdhrdc.z03.azurefd[.]net
  • heemawarii-h6cdedfrfxdcgbbp.z03.azurefd[.]net
  • hyujryhty-b2gtgwd9dzb2b2a7.z02.azurefd[.]net
  • looolopine-gwdubzevcvb7hmff.z03.azurefd[.]net

This campaign uses a fake Paperless Post invitation to push targeted personnel at a Kentucky organization into signing in with a webmail account, and it is built on the Office360 modal-form kit. The click was detected on October 7, 2026. The page asks the user to pick Outlook, Office365, Yahoo, AOL or other mail, and each button opens the same modal. The modal posts the email address and password by AJAX to processmail.php. The page then polls check.php, and the operator's reply decides the next step: a wrong-password message, an OTP prompt, a phone number prompt, or a redirect. The OTP and phone values go by form POST to process.php, with a five-minute countdown to add urgency. The Gmail button is a plain link to a separate ./google page. Cloudflare scripts and a hidden iframe from the GoGuardian extension host are also present on the page. Because the operator answers each submission live, the kit can relay one-time codes before they expire. Defenders should expect MFA codes to be captured on pages like this one.

This campaign impersonates a Microsoft secure file-sharing notice, offering a 56.1KB PDF behind an identity check, and it targeted a Kentucky organization. The click was detected on October 6, 2026. The page is a multi-step login kit built from hidden sections that script swaps in and out. The first section asks for a work email and a password. Later sections mimic the Microsoft MFA picker with Authenticator approval, a verification code, SMS and a phone call. The inputs are not wrapped in a form. The buttons call JavaScript functions such as submitPass() and OfficeSendVerify(), and OTP codes go out through OfficeAjaxVerify(), so the stolen values can be relayed while the victim is still signing in. The email field is read-only and the password field and button start nearly transparent. CSS also blocks text selection. A hidden GoDaddy Microsoft 365 sign-in section sits in the same page, and the page carries unrendered PHP template tags for token, email and key. Because the kit collects the one-time code live, a password reset alone may not undo the compromise, and detection should key on the page structure instead of the brand shown.

This campaign impersonates the Microsoft account sign-in page and asks a targeted Georgia organization's users to enter a password for an account whose address is already filled in. Three detections occurred on October 3, October 4, and October 5, 2026. The page copies the Microsoft Fluent sign-in layout and loads its logo and styling assets from Microsoft's own login CDN. The victim's address is pre-filled in a hidden input and shown in the identity banner under the heading "Enter your password". The password input is named passwd and sits outside any working form. The only form on the page is an empty hidden POST form. The Next button calls a script function named redlogin() rather than submitting. The page also loads the Socket.IO client library from cdnjs, which points to a real-time channel back to the operator instead of a plain form POST. A browser-challenge meta tag and a robots tag set to noindex and nofollow keep the page out of crawlers. The campaign used a long URL path made of random word-and-number tokens on a hostname under bralumekio[.]com. Because no form action exists, detection that looks for a POST to an attacker domain will miss this page. Socket.IO traffic from a page styled as a Microsoft sign-in is a better signal.

This campaign impersonates a Microsoft sign-in page and targeted a Kentucky organization with a replica of the Microsoft converged sign-in flow. Two detections occurred on October 2 and October 4, 2026. The page is a full copy of the Microsoft ConvergedSignIn page, including its $Config block, its Knockout templates and its footer links to Microsoft terms and privacy pages. Credentials are submitted to the urlPost value https://login.rebrand.org.nz/4GGfkj/jI6be, which sits on the attacker's own domain. The flow has fields for email, password, phone number and a one-time code, so the operator can collect MFA codes as the victim types them. Several measures keep scanners away from the page: crawler user-agent blocking, a CAPTCHA, text selection disabled, a delayed redirect and a periodic check loop. Visitors with JavaScript off are sent by a noscript meta refresh to a jsdisabled path on the same host. The page also loads a hidden iframe from portal[.]rebrand[.]org[.]nz and spreads its static assets over randomly named subdomains of rebrand[.]org[.]nz. Defenders should treat any Microsoft-looking login on a non-Microsoft domain as hostile, and domain-based blocking on the sign-in host will catch more than blocking on page content.

This campaign impersonates a Microsoft account sign-in page and asks a Kentucky organization's users to re-enter a password for an account whose address is already shown. The click was detected on October 1, 2026. The page is the second step of a Microsoft login flow. The victim's address is pre-filled in a hidden input and displayed above an Enter password prompt. The password input is not wrapped in a form. The Sign in button calls a JavaScript function, redlogin(), instead of submitting to a form action. The page loads the Socket.IO client from cdnjs, which points to a real-time channel for sending the entry out, though the receiving endpoint is not visible in the static page. A browser-challenge meta tag and a CAPTCHA step gate the page, and some content is Base64 encoded. Links such as Terms of use and Forgotten my password point to #. The long underscore-delimited path on a throwaway subdomain acts as a per-victim token. Detection that depends on form action URLs will miss this pattern, so it needs to look at script behavior and WebSocket traffic.

This campaign impersonates a Microsoft password prompt and targeted a Kentucky organization, with the victim's address already filled in on the page. The click was detected on September 30, 2026. The page is a single Microsoft-styled screen titled "Sign in to your account" with the heading "Enter password". The victim's address is pre-filled in a hidden input and shown as the account name. The password input is not wrapped in a form. The Sign in button is a plain button that calls a JavaScript function named redlogin(), so there is no form POST for a scanner to find. The page loads the Socket.IO client library from cdnjs, which points to a real-time channel back to the operator rather than a standard submission endpoint. Logos, background and icons load from relative paths on the same host. The long underscore-delimited path of random words makes each link unique. Detection that looks only for a form action will miss this page. Defenders should watch for a Microsoft sign-in look on a non-Microsoft domain that opens a WebSocket connection.

This campaign imitates a Microsoft account sign-in page and asks for a password and a phone number, and it targeted a Minnesota organization. The click was detected on September 28, 2026. The page is a basic form POST. The password field and a phone number field sit inside one form that submits to https://anself.com/SB7921/secure/signin/process, on the same domain that serves the page. The victim's address is pre-filled in a button above the heading, so the page reads as the second step of a Microsoft sign-in. The phone field uses a jQuery Mask plugin to force a US format of (000) 000-0000, which keeps the harvested number clean. Page assets such as the logo and favicon load from a path on the same host, while Bootstrap and jQuery come from public CDNs. The Forgot password and Create an account links point to # and go nowhere. Asking for a phone number next to the password gives the operator a second identifier for follow-up contact or account recovery abuse. Detection should key on credential forms that post to a path on a host unrelated to Microsoft, because the page carries no evasion code and no kit signature.

This campaign impersonates the Microsoft sign-in page and targeted a Minnesota organization with a link that opens a cloned converged sign-in flow. The click was detected on September 28, 2026. The page is a copy of the Microsoft ConvergedSignIn page. Its configuration points the credential post, urlPost and urlPostAad, to https://login.securedistrict[.]icu/6GDChj/GHbYE, so the submitted values go to the attacker's host. The page asks for an email address, then a password, and it also carries fields for a phone number and an MFA or OTP code. Every other config URL, including the GetCredentialType and one-time-code endpoints, is rewritten to the same lookalike domain. A hidden zero-size iframe loads portal.securedistrict[.]icu/Prefetch/Prefetch.aspx. The page also uses a CAPTCHA step, a noscript meta refresh, delayed redirects and a periodic check loop to filter out crawlers. Part of the script is base64 encoded and high-entropy. Because the markup is a faithful copy, a visual check of the page will not separate it from the real sign-in. Detection has to rest on the domain and on where the post goes.

This campaign used a fake AOL sign-in page titled "Secure Login Portal" to collect passwords from a Florida organization. The click was detected on September 28, 2026. The page shows the AOL name over a promo panel for a faster PC, which does not match a real AOL sign-in. The username step is hidden, and the password step is shown with the victim's address pre-filled in a badge. The password input is not inside a form. The Sign In button calls submitLogin(), which only checks that the field is not empty and then sends the browser to blocked.html. The script makes no network request, so the typed password stays in the browser on this page. Any capture would have to happen on the next page or by another route that the page itself does not show. The host is a smtp-prefixed subdomain of aybemail[.]com, which suggests the link was meant to look like a mail service. Defenders should treat the pre-filled address and the hand-off to a second page as the signals to follow, and should inspect blocked.html and what it asks for next.

This campaign impersonates the Amazon sign-in page with a Prime re-authentication pretext and targeted a Kentucky organization. The click was detected on September 28, 2026. The page is a basic form POST. The form submits the email address and password to /signin/process on the same host. It is a two-step clone. The email field shows first, and a script then swaps in the password panel and echoes the typed address into it. The email is checked with a regular expression, and the password must be at least six characters. A hidden error box reads "We cannot find an account with that email address" and appears when the address fails the check. The Amazon logo, footer text and class names are copied from the real page. The assets load from a /arahmataAngin/ path on the attacker's host. The hostname is a long, random-looking subdomain chain that begins with prime-reauth.document and ends in doc8fs743h26ash74724h6[.]online. Defenders can match on the Amazon page title served from a non-Amazon domain. The same-host POST to /signin/process is a second signal.

This campaign impersonates the Google Accounts password step, showing a "Too many failed attempts" notice, and it targeted a Kentucky organization. Two detections occurred on September 27 and September 28, 2026. The page is a copy of Google's lite sign-in password screen with the victim's address pre-filled in a hidden identifier field. A visible password input named Passwd sits inside a form that POSTs to a relative /v3/signin/challenge/pwd path on the attacker's own host, so the server behind the page receives the credentials. The page loads a botguard.js file from a Replit-hosted origin and carries a large block of copied Google script that adds noise to the source. Two scripts screen visitors first. One scores the browser for automation and headless signals and sends reports to a /.detect endpoint. The other blocks right-click and view-source and redirects to google[.]com when DevTools shortcuts are used. The campaign rotated across 2 hostnames on yuansenmenye[.]com. Because the form posts to a relative path, the destination is the phishing host itself. Detection should therefore key on the Google-branded page served from an unrelated domain rather than on an external exfiltration address.

Related subdomain variants:

  • 3x8rtih43c.yuansenmenye[.]com
  • q0u5het4clerwq4.yuansenmenye[.]com

This campaign impersonates a Microsoft sign-in page, with a hidden GoDaddy-branded Microsoft 365 login behind it, and it targeted an Illinois organization. Both detections occurred on September 25, 2026. The page is a multi-step flow built from section_* and inp_* elements, and the inputs are not wrapped in a submitting form. Buttons call JavaScript functions such as submitPass(), checkEmail() and OfficeAjaxVerify(), and the SOC's decompiler traced submission to a mail/smtp relay. The flow collects the email address and password, then moves to a fake MFA picker. That picker offers Authenticator approval, a verification code, a text message or a phone call, and it asks for a phone number. A code entry screen follows, so a one-time code can be captured as well as the password. The error text on the password step is split with hidden spans, which breaks up phrase matching in page text. CSS also disables text selection. The Microsoft logo is loaded from Microsoft's own CDN, so the brand assets are genuine. A defender should expect the MFA step to be harvested in the same session, and a code typed into a page like this should be treated as exposed.

This campaign uses a fake online invitation page that asks users to sign in with their email provider, and it targeted a Minnesota organization. The click was detected on September 25, 2026. The page tells the visitor to pick Outlook, Office365, Yahoo, AOL or other mail to view an invitation. Each button opens a modal with an email and password form. jQuery intercepts the submit and sends the fields by AJAX POST to processmail.php. The page then shows "Incorrect Password", so the victim is likely to type the password again. A second modal says an OTP will arrive in 1 to 5 minutes and shows a five-minute countdown. The OTP form posts to process.php, which lets the operator collect a live code while it is still valid. The Gmail button is a plain link to another path on the same domain, and the page matches an Office360 modal-form kit. Page branding is a mix of invitation-service and mail-provider logos, so detection based on a single brand will miss it. Defenders should treat the password-then-OTP sequence on an unfamiliar host as a signal that one-time codes are being relayed.

A fake Outlook Web App sign-in page targeted a Minnesota organization with a link that carries the recipient's address in an encoded URL parameter. The click was detected on September 24, 2026. The page is a single login screen titled Outlook with a Domain/Username field and a Password field. A script reads the mxid parameter, decodes it with atob() and pre-fills the address field. On submit, the script cancels the normal form action. It builds a text message with the attempt number, address and password. It sends that message by POST to owasend_telegram.php, a server-side script that relays the data to Telegram. The first attempt shows an incorrect password message and clears the password field, so the victim types the password a second time. The second attempt redirects to the real office.com sign-in page. The findings flag email-based exfiltration, but the HTML shows the Telegram relay, so a mail-only rule would miss the real destination. Defenders can look for a POST to a local PHP script from a page that claims to be Outlook. The double entry also gives the operator two password samples per victim.

This campaign is a fake Windows security alert that tells a user in a Kentucky organization the device is locked and pushes them to call a toll-free support number. The click was detected on September 24, 2026. The page does not capture credentials. The username and password inputs are marked readonly and are not wrapped in a form. The Login, Cancel, Run Scan and Decline buttons are all links to "#", so nothing is submitted anywhere. Red text names a fake threat, system-spawn.dll, and says access has been blocked. A marquee and a Windows Support Center block borrow Microsoft branding to look credible. The page is hosted on an Azure Container Apps hostname. Because there is no form post or script to flag, detection has to rely on the URL, the hosting pattern and the page text. Users are steered to a live phone call, where the actual theft or remote-access request would happen.

Recommendations

  • Block the attacker-registered credential hosts at the proxy and DNS layer: bralumekio[.]com, jentaxilon[.]com, retroprobe[.]com, rebrand.org[.]nz, securedistrict[.]icu, anself[.]com, morninggoldenlight[.]de, silverveil[.]org, ptiry[.]click, info2party[.]com, yuansenmenye[.]com, doc8fs743h26ash74724h6[.]online and aybemail[.]com. Their randomized subdomains and per-victim token paths change from link to link, so block at the registered domain rather than the full hostname.
  • Do not block the Azure Static Web Apps, Azure Front Door, Azure Container Apps or b-cdn[.]net platforms that hosted the scareware pages. Detect on page content instead: a toll-free number repeated across a SmartScreen-style box, a chat widget and a marquee; read-only login inputs with no form; and strings such as "Memory access violation at 0x88412" or system-crash.dll.
  • Two campaigns also passed the number in the URL query string, so a URL match on a phone-number parameter is a second detection point.
  • For the Microsoft password re-entry pages that relay over Socket.IO, alert when a page with Microsoft sign-in styling on a non-Microsoft domain loads the Socket.IO client from cdnjs or opens a WebSocket. Add page-source rules for the redlogin() handler and for the SectionKit section_*/inp_* structure with submitPass() and OfficeAjaxVerify(), because these pages have no form action for a POST-based rule to match.
  • For the cloned Microsoft ConvergedSignIn pages, inspect the urlPost and urlPostAad values in the $Config block and alert when they point anywhere other than Microsoft. Apply the same check to the Google password clone posting Passwd to /v3/signin/challenge/pwd and to the Amazon clone posting to /signin/process on an unrelated host.
  • For the OTP-relay invitation lures built on the Office360 modal-form kit, hunt for an AJAX POST to processmail.php followed by polling of check.php and a POST to process.php. Treat any account whose owner typed a one-time code into one of these pages as fully compromised, since the operator could replay the code before it expired.
  • Tell users that a browser page showing a locked device and a support phone number is never genuine, and that the correct response is to close the tab, not to call. Also tell them that a sign-in page with their address already filled in, an "Incorrect Password" message followed by a code prompt, or a party invitation that asks for a mailbox login should be reported before they type anything.

Learn how PhishID can help protect your district from cutting-edge threats like these! Schedule a Demo