Skip to content
nulltapE-reader edition

← All e-reader articles

4 min read

Huntress Traces Token Replay Across 97 Microsoft 365 Accounts

A replayed sign-in session reached accounts without endpoint evidence; Huntress tied Defender incidents to identities and found rogue authentication methods and hidden mailbox replies.

Read on the standard site

An incoming email reply diverted from the main inbox into a shadowed side compartment.

A stolen Microsoft 365 session let attackers enter accounts without touching a customer endpoint. In a campaign tracked from July 14 to September 22, 2026, Huntress says 97 accounts at 79 organizations were compromised. Attackers registered their own devices and authenticator apps, then used inbox rules to hide mail from account owners. For a team working an identity alert, a clean workstation is only part of the picture.

Huntress has expanded its Microsoft Defender integration to bring incidents from Defender for Identity, Defender for Office 365, Defender for Cloud Apps, and Entra ID Protection into its managed response queue. It now attaches an incident to the identities it names, even if the incident contains no device. Huntress says tenants with Microsoft Defender for Endpoint connected automatically receive the additional Defender incident data.

A sign-in without a workstation

The campaign began with a Microsoft alert supported by Microsoft Threat Intelligence Center intelligence, according to Huntress. One technique was adversary-in-the-middle phishing: a fake sign-in page relayed a login to Microsoft, captured the session token issued after MFA, and let the attacker replay that token to reach the account. The replayed session gave attackers account access without compromising a customer endpoint. Huntress says its Managed ITDR detections first identified the activity; Entra ID Protection separately flagged the same attacker infrastructure across multiple customer environments.

Once inside, the attackers registered rogue devices and their own authenticator apps to retain access. They also created inbox rules that concealed messages. In one case, an attacker used a hijacked mailbox to send a payroll-related email to an outside accountant and hid the accountant’s replies from the mailbox owner. Check inbox rules for hidden replies and trace any exchange the account owner missed.

Why the incident has no device

Microsoft describes a Defender incident as a container for related alerts, including lists of impacted users and devices. A user can be the affected entity even when no endpoint alert joins the incident. In Huntress’s recent two-week sample, 85% of medium- and high-severity Defender incidents had no device-related evidence. That percentage describes incidents in that sample, not the proportion of compromised accounts.

Huntress says its earlier pipeline relied more heavily on endpoint context. Its new routing associates each Defender incident with involved endpoints, protected identities, or both, and signals every affected account when an incident names several. “This campaign is the reason we rebuilt the pipeline,” the company wrote. Its analysts can also use Microsoft Defender advanced hunting to look back as far as 30 days in a customer’s tenant, following an account’s sign-ins and subsequent activity or connecting a credential-theft alert to a process tree. Huntress says it applies per-tenant query limits to preserve the Microsoft hunting quota shared with other tools.

Follow the account after containment

Start with the affected identities in Defender’s incident view. For an account suspected of token replay, examine its sign-in history, registered devices, authentication methods, and mailbox rules. In this campaign those artifacts connected the initial access to persistence and concealed mail; a device-only investigation would miss the account path. Huntress recommends revoking the user’s sessions, resetting the password, and then reviewing devices and MFA methods in Entra ID. Remove unauthorized registrations and rules through your normal identity and mail-response workflow, and investigate the hidden conversation and any other accounts named in the same incident.

Microsoft’s Entra guidance gives administrators a way to block new sign-ins and revoke sessions from the user’s Overview page. It also explains the timing limit: existing access tokens can remain valid until they expire, and an application that issued its own session cookie controls when that cookie ends. If the attacker reached another application, have its owner terminate that application’s session too. A password reset on its own leaves attacker-registered devices and MFA methods available, as Huntress warns.

Check the identity and mailbox

After containment, verify the affected user’s Entra device and MFA-method lists against approved registrations, and confirm that the inbox rules no longer divert replies. Recheck the account’s sign-in activity and the full Defender incident for other named identities; extend the response to those accounts where the evidence warrants it. For applications with their own sessions, confirm the application-side session was terminated. Finish by checking the account’s registrations, sessions and mailbox behavior against the owner’s intended state.

Primary sources