Manipulate tenant branding fields
Within Microsoft Entra ID, tenants have customizable branding information such as the organization name. Researchers found attackers placing scam text directly in these fields—for example, fake purchase alerts or instructions to call a support number.
Trigger a system notification
The attacker then initiates a legitimate Microsoft workflow, such as adding a new email sign‑in or recovery method for a user in the tenant. Instead of their own address, they enter the victim’s email address.
Microsoft sends the message
That action causes Microsoft’s notification system to send an automated email to the target address. Because the email template includes the tenant name or branding fields, the attacker’s scam text appears inside what is otherwise a legitimate Microsoft-generated email.
In effect, the attacker injects a phishing message into a real Microsoft notification template.
Researchers have observed several tactics used to make the messages more convincing and harder for filters to detect.
• Subject-line hijacking – Attackers pack long scam messages into branding fields so the fraudulent text dominates the subject line or preview text.
• Character substitutions – Look‑alike characters and homoglyphs can evade keyword detection systems.
• Phone-number obfuscation – Numbers may be written with letters or unusual characters to bypass automated spam filters.
These tricks help the emails survive automated scanning while still conveying the scam message to humans.
The most dangerous aspect of this campaign is that the messages are not simply spoofed.
Because Microsoft’s own systems generate and send the notifications, the emails can pass standard authentication checks such as SPF, DKIM, and DMARC, which many mail systems rely on to verify senders.
Recipients also see a familiar Microsoft sender address associated with account security alerts, which increases trust even if the message content is suspicious.
This combination—trusted sender infrastructure plus injected scam text—makes the emails far more convincing than typical phishing attempts.
Investigations by journalists and security researchers indicate the abuse has been active for months.
Reports describe:
• Multiple phishing emails sent from the Microsoft notification address across different inboxes.
• Scam themes including fake purchase alerts, billing problems, and urgent support requests.
• Large numbers of disposable Microsoft 365 tenants being used and discarded in a "burn‑and‑churn" pattern.
Anti‑spam researchers have criticized the level of customization allowed in automated notification workflows, warning that trusted platform infrastructure can be turned into a phishing delivery channel if safeguards are weak.
Importantly, reporting has noted that Microsoft had not publicly confirmed the exact technical root cause at the time these incidents were documented.
Because the sender address can be legitimate, users should rely on context and behavior rather than the “From” field alone.
Warning signs include:
• Unexpected purchase or fraud alerts you did not initiate
• Messages urging you to call a phone number immediately
• Requests to click links or resolve urgent billing issues
• Verification codes for actions you never requested
Even a real Microsoft email address does not guarantee the message content is trustworthy.
If you receive a suspicious Microsoft notification email:
• Do not click links or call numbers provided in the message.
• Open Microsoft services manually by typing the official website into your browser.
• Check your account activity directly from your account dashboard.
• Report the message as phishing in your email client or to your organization’s security team.
• If you interacted with the message, change your Microsoft password and review sign‑in activity.
This incident highlights a growing trend in cybercrime: attackers increasingly exploit trusted platforms and legitimate infrastructure instead of relying solely on fake domains. When a real service delivers the phishing message, traditional trust signals are no longer enough to keep users safe.
Comments
0 comments