
Russia's GRU-linked APT28, which Microsoft tracks as Forest Blizzard, has been taking over small-office and home routers and pointing their DNS at attacker-controlled resolvers. That let it watch, and sometimes intercept, connections to Microsoft 365 and government services. Microsoft published details on 7 April 2026, and the same day reports said authorities and private partners had disrupted the infrastructure. For engineering teams with remote staff, it is a reminder that the network a laptop joins is part of the attack surface.
What happened
Microsoft Threat Intelligence attributes the campaign to Forest Blizzard and a sub-group it tracks as Storm-2754. It has seen the activity since at least August 2025. The actor exploited vulnerable SOHO devices and changed their default network settings so that connected clients used actor-controlled DNS resolvers. Microsoft assesses that those resolvers almost certainly ran dnsmasq on port 53. It also found the altered DNS settings had reached Windows machines behind the compromised routers.
Microsoft says the malicious DNS infrastructure affected more than 200 organisations and about 5,000 consumer devices. Most DNS queries were transparently proxied, so users noticed nothing. For selected targets, the actor ran adversary-in-the-middle (AiTM) attacks on TLS connections to a subset of Outlook on the web domains, and on non-Microsoft servers belonging to at least three African government organisations. In a limited number of cases, the actor spoofed DNS answers and sent victims to servers presenting an invalid TLS certificate impersonating the Microsoft service. Users who clicked through the warning exposed their traffic. Microsoft found no compromise of its own assets.
The original reporting described a coordinated takedown by law enforcement and industry partners, and named the infrastructure "FrostArmada" (Lumen's Black Lotus Labs has published research under that name). We could not read the law-enforcement announcement, so treat the takedown details as reported.
Who is affected
- Owners of vulnerable, internet-exposed SOHO routers, including remote workers, small offices and branch sites using consumer-grade gear.
- Organisations in government, IT, telecommunications and energy, the sectors Microsoft lists.
- Microsoft 365 users who signed in to Outlook on the web from a compromised network, especially anyone who accepted a certificate warning.
What it means for developers and engineering teams
DNS hijacking at the router is quiet and powerful. Changing the DHCP-advertised resolver on one device redirects every client behind it, and nothing on the endpoint has to be compromised. Most of the time TLS still protects users: a hijacked name resolves to an attacker server that can't present a valid certificate for outlook.office.com. The attack depends on users clicking through warnings, or on apps that don't validate certificates properly.
So for developers, this comes down to TLS hygiene. Never ship clients that disable certificate validation "temporarily". Use HSTS (and preload where you can) on login domains so browsers refuse click-through. Consider certificate pinning or Certificate Transparency monitoring for high-value internal apps. On the identity side, phishing-resistant authentication such as passkeys or FIDO2 binds the credential to the real origin, so a proxied login page can't replay it. TOTP and push approvals don't protect you this way.
Finally, cleanup has limits. Fixing a router's DNS stops new interception, but any credentials or session tokens already stolen remain valid until you revoke them.
What to do now
- Check resolver settings on endpoints and routers. On Windows:
Get-DnsClientServerAddress. On macOS/Linux:scutil --dnsorresolvectl status. Flag resolvers you don't recognise. - Patch or replace SOHO routers, disable remote administration, and change default credentials. Don't use home-grade routers in corporate offices.
- Enforce encrypted or managed DNS on corporate endpoints, for example DNS-over-HTTPS to a known resolver or Zero Trust DNS on Windows, so a hostile DHCP resolver is ignored.
- Move privileged users to passkeys or FIDO2, and restrict where MFA methods can be registered using Conditional Access.
- Hunt and revoke. Review Entra ID risky sign-ins and threat-intelligence risk events. For suspected victims, revoke sessions and refresh tokens. A password reset alone is not enough.
- Turn on HSTS for your own login and admin domains:
Strict-Transport-Security: max-age=31536000; includeSubDomains; preload.