Back to Articles

EASM

Compromised Mailboxes Are Among Italy's Top Threats: How to Harden Microsoft 365 and Google Workspace

13 min read

ACN's August 2026 summary lists mailbox compromise among the most frequent threats and valid credentials as the second way in. How mailboxes are taken over, what attackers do first, and a practical hardening checklist for Microsoft 365 and Google Workspace.

microsoft-365
google-workspace

At the end of September 2026 the Italian National Cybersecurity Agency (ACN) published its operational summary for August. As reported by ICT Security Magazine, compromised mailboxes were among the most frequent threats of the month, with 47 cases, and the prevalent mode in the telecommunications sector. Intrusions with valid credentials, 42 cases, were the second initial access vector after the exploitation of vulnerabilities. A company's email tenant is part of its attack surface just as much as its firewall, which is why this guide sits next to our guide to external attack surface management. Below: how mailboxes are taken over in 2026, what attackers do in the first hour, and the settings to check in Microsoft 365 and Google Workspace.

Key Takeaways

  • MFA alone no longer settles it. Adversary-in-the-middle kits and device code phishing get around one-time codes and push approvals.
  • The first move after a takeover is quiet: an inbox rule, an external forward, a delegate or an OAuth app that survives a password reset.
  • Most of the fix is configuration: phishing-resistant MFA for admins, conditional access, no legacy authentication, controlled app consent and guest access, no automatic external forwarding.
  • Logs are short by default. Decide now where sign-in and audit logs go and for how long.
  • Check posture continuously, because settings drift as admins, apps and guests come and go.

How mailboxes get taken over in 2026

Adversary-in-the-middle (AiTM) phishing

The victim lands on a page that proxies the real Microsoft or Google login. They type their password and approve MFA, and the proxy captures the session cookie that the identity provider issues afterwards. The attacker replays that cookie and is in, without needing the password or the second factor again. The ENISA Threat Landscape 2026 reports that Tycoon 2FA, an AiTM kit designed to bypass multi-factor authentication, was reportedly the most widely observed phishing kit worldwide in 2025.

Device code phishing

Here there is no fake page at all. The attacker starts a legitimate sign-in for a device that "cannot type", receives a short code and persuades the victim to enter it on the real Microsoft page. The victim completes MFA on a genuine site, and the attacker's session receives the tokens. ENISA notes that multiple intrusion sets used this technique in 2025, often after first contact on Signal or WhatsApp. We cover it in detail in our article on device code phishing.

Password spraying and stolen credentials

Attackers try a few common passwords across many accounts, looking for one without MFA or reachable through a protocol that does not enforce it. Credentials stolen by infostealer malware on a home PC are the other source, and they often come with valid session cookies. Our guide to credential exposure explains how these leaks happen.

Consent phishing

Instead of a password, the victim is asked to approve an application that requests permission to read mail or files. Once granted, the app's access does not depend on the user's password and survives a reset.

What attackers do in the first hour

A compromised mailbox is rarely used to send spam straight away. The goal is usually invoice fraud, data theft or a foothold for the next phishing wave, so the first steps are designed to stay unnoticed:

  • Inbox rules that move or delete messages containing words such as "invoice", "payment", "IBAN" or "bonifico", or the names of the finance team, so the victim does not see replies.
  • External forwarding of all mail, or of selected messages, to an address the attacker controls.
  • Delegation of the mailbox to another compromised account.
  • A new MFA method registered to the attacker's own phone or authenticator.
  • An OAuth app granted mail access, which keeps working after the password is changed.
  • Searches of the mailbox and of SharePoint, OneDrive or Drive for invoices, contracts and credentials.

Every item on this list is a configuration change that can be prevented or detected. That is the logic of the checklist that follows.

Checklist part one: identity and sign-in

MFA, and phishing-resistant MFA for administrators

  • Require MFA for every user, with no permanent exclusions except documented break-glass accounts.
  • Give Global Administrators, Exchange and SharePoint admins, and Google Workspace super admins phishing-resistant methods: FIDO2 security keys or passkeys. Codes and push approvals are what AiTM kits are built to relay.
  • Keep administrator accounts separate from daily mailbox accounts, and keep their number small.

Conditional access (Microsoft 365) and context-aware access (Google Workspace)

  • In Microsoft Entra ID, build conditional access policies that require MFA, require compliant or hybrid-joined devices for administrators, and block sign-ins from countries where you do not operate.
  • Restrict the device code authentication flow to the users and devices that actually need it, typically meeting-room devices, and block it for everyone else.
  • Where your Google Workspace edition supports it, use context-aware access to limit admin and sensitive app access to managed devices.
  • If you do not have the licences for conditional access, at least enable Microsoft's security defaults.

Block legacy authentication

Protocols that do not support modern authentication cannot enforce MFA, which is why password spraying targets them. Block legacy authentication with a conditional access policy, disable it per mailbox protocol where it is not needed, and in Google Workspace make sure access for less secure apps is off. Check sign-in logs for any remaining legacy clients before you close the door, so a printer or a scanner does not stop working on Monday morning.

If you still run AD FS

Hybrid identity widens the target. On 14 July 2026 CISA added CVE-2026-56155, a privilege elevation vulnerability in Microsoft Active Directory Federation Services, to its Known Exploited Vulnerabilities catalog. If you still federate Microsoft 365 through AD FS, patch it, and consider Microsoft's own decommissioning guidance for moving to cloud authentication.

Checklist part two: apps, guests and delegation

OAuth app consent

  • In Entra ID, stop users from consenting to apps that request access to mail or files. Allow consent only for verified publishers and low-risk permissions, and turn on the admin consent workflow so requests reach IT instead of being refused silently.
  • Review existing enterprise applications. Look for apps with mail read or send permissions and offline access, unverified publishers, and apps nobody recognises.
  • In Google Workspace, use API controls to mark third-party apps as trusted, limited or blocked, and review which apps have access to Gmail and Drive scopes.

Guests

  • Limit who can invite guests and restrict what guests can see in the directory.
  • Review guest accounts every quarter and remove those tied to finished projects.

Forwarding and delegation

  • In Exchange Online, set automatic forwarding to external domains to off in the outbound anti-spam policy, and allow exceptions only for named mailboxes with a business reason.
  • In Google Workspace, disable automatic forwarding for users who do not need it, and control whether users can delegate their mailbox.
  • List existing forwarding rules and delegates now. In incident response work, an old external forward is one of the most common findings.

Sharing in SharePoint, OneDrive and Drive

Mail is only half of the tenant. The documents an attacker searches for live in SharePoint, OneDrive and Drive.

  • Avoid "anyone with the link" sharing for the whole tenant. If you need it, set an expiry and limit it to specific sites.
  • Make "specific people" the default link type.
  • In Google Drive, restrict sharing outside the domain or limit it to allowlisted partner domains, and warn users when they share externally.
  • Review externally shared files on finance, HR and legal sites first.

Email authentication: SPF, DKIM and DMARC

A compromised mailbox inside your tenant is one problem; someone sending convincing mail in your name from outside is another. Publish SPF, sign outgoing mail with DKIM and move DMARC from monitoring to quarantine or reject once your legitimate senders are aligned. These records are public, so anyone can check them, including attackers looking for a domain to spoof.

Monitoring sign-ins and changes

  • Keep the logs. Make sure the Microsoft 365 unified audit log is on, and know how long sign-in and audit logs are kept with your licence: by default it is a matter of days or weeks. Export them to a SIEM or log store if you need longer.
  • Alert on the changes listed above: new inbox rules with external forwarding, new delegates, new MFA methods, new OAuth consents and admin role assignments.
  • Watch for impossible travel, MFA fatigue and token reuse from unfamiliar locations.
  • In Google Workspace, use the alert center and the login and admin audit logs for the same events.

If a mailbox is already compromised

  1. Revoke the user's sessions and refresh tokens, then reset the password. A reset alone does not end a stolen session.
  2. Remove MFA methods the user does not recognise.
  3. Delete malicious inbox rules, external forwards and delegates.
  4. Revoke OAuth grants to unknown apps.
  5. Check sent items and inform counterparties who may have received fraudulent payment instructions.
  6. Assess whether NIS2 or GDPR notification duties apply, within their deadlines.

How RECON Cloud helps

RECON Cloud connects to Microsoft 365 and Google Workspace as SaaS connectors with read-only access. For Microsoft 365, a Global Administrator approves read-only Microsoft Graph permissions once: directory, policies, app registrations, mailbox settings and SharePoint sharing. For Google Workspace, a service account with domain-wide delegation reads users, groups, 2-Step Verification, Gmail forwarding and delegation settings and Drive sharing metadata. RECON never reads the contents of mail or files, and access is revoked from your own admin console.

Each scan checks MFA and 2-Step Verification for admins, conditional access, legacy authentication, guest and consent settings, risky or unverified OAuth apps, external mail forwarding and delegation, anonymous or external sharing in SharePoint, OneDrive and Drive, and the SPF, DKIM and DMARC records of your domains. Findings are mapped to compliance controls, including the CIS Microsoft 365 benchmark and the NIS2 Article 21(2) measures. Because configuration drifts, scheduled scans show what changed since the last one.

Technology closes most doors; people decide whether to type a code on the wrong page. Our Human Firewall programme trains staff on exactly these lures.

Next steps

  1. Enforce phishing-resistant MFA for every administrator this month.
  2. Block legacy authentication and restrict the device code flow.
  3. Turn off automatic external forwarding and list existing forwards and delegates.
  4. Restrict user consent to apps and review the apps already granted.
  5. Set sharing defaults to specific people and review external shares on sensitive sites.
  6. Decide where sign-in and audit logs go, and for how long.
  7. Move DMARC towards enforcement.
  8. Schedule a posture check so you find the next drift before an attacker does.

If you want to see where your tenant stands today, connect it to RECON Cloud with read-only access: the first scan shows which of these settings are already right and which are not, with the evidence for each.

microsoft-365
google-workspace
email-security
mfa
conditional-access
sspm