Microsoft Outlook Add-in Stolen 4,000 Microsoft account Credentials and Credit Card Numbers

In Cybersecurity News - Original News Source is cybersecuritynews.com by Blog Writer

Microsoft Outlook Add-in Stolen Credentials

Security researchers have identified the first documented instance of a malicious Microsoft Outlook add-in being used against users in real-world scenarios.

A compromised meeting scheduler named AgreeTo was used to steal over 4,000 Microsoft account credentials, credit card numbers, and answers to banking security questions.

AgreeTo began as a legitimate open-source project published to the Microsoft Office Add-in Store in December 2022.

Koidex report for AgreeTo add-in (Source: Koi AI)

It was a functional meeting tool with positive reviews. However, the developer eventually abandoned the project and deleted the associated Vercel deployment.

This left the add-in’s hosting URL (outlook-one.vercel.app) orphaned and available for registration.​ An attacker claimed the available URL and deployed a phishing kit.

AgreeTo page on the Office Add-ins marketplace (Source: Koi AI)

Because the add-in was never removed from Microsoft’s store, the attacker’s phishing page was served directly in the trusted Outlook sidebar to anyone who still had the add-in installed or had downloaded it fresh.​

The Technical Flaw: Remote Dynamic Dependencies

The attack exploited the architecture of Office add-ins. Unlike traditional installed software, add-ins are “remote dynamic dependencies,” essentially XML manifests that load a URL in an iframe.​

Stage Description
The Gap Microsoft reviews the manifest only during the initial submission (2022 in this case).
The Exploit Microsoft does not continuously verify the live content hosted at the approved URL.
The Result The attacker swapped the scheduling tool with a fake Microsoft login page without triggering a new security review.

When victims opened AgreeTo, they were prompted to log in. A script harvested their credentials and IP addresses, sending the data to the attacker via a Telegram bot.

Add-in Displays Fake Login (Source: Koi AI)

​Koi Security discovered the campaign after identifying the attacker’s poorly secured exfiltration channel.

They recovered the full dataset of 4,000 victims, which included banking details and security answers targeting Canadian institutions.

The add-in had “ReadWriteItem” permissions, which technically allowed the attacker to read and modify users’ emails. However, the attack primarily focused on credential harvesting.​

Microsoft has removed the add-in following the report. This incident highlights a critical supply chain risk: trusted software can silently become malicious years after approval if its infrastructure is abandoned.

Follow us on Google News, LinkedIn, and X for daily cybersecurity updates. Contact us to feature your stories.