When an attacker holds your email, chat or identity provider, every incident call, ticket and Teams or Slack thread about the response is also an intelligence feed for the attacker. CISA has documented a group that searched victims' Slack, Microsoft Teams and Exchange for conversations about the intrusion and joined incident response calls, and Microsoft found that Midnight Blizzard searched senior leadership and security mailboxes for information about itself. Out-of-band (OOB) communications are channels that do not share identity, devices or infrastructure with the compromised estate, provisioned and tested before they are needed. This guide covers threat assumptions, channel options, identity and device design for the OOB environment, activation triggers, operating rules, testing and metrics, for incident response leads, CISOs and the teams who will build the channel. It complements the Crisis Management Guide, which covers the crisis structure around it.
Key takeaways
- Assume the attacker reads everything on the corporate stack. If the intrusion touches the identity provider, email, chat, ticketing and video conferencing are all suspect, including calendar invites and dial-in numbers.
- Independence is about identity, not just the app. A second chat tool that signs in through the same identity provider, on the same managed laptops, is not out of band.
- Pre-provision and pre-enrol. Accounts, devices, hardware keys, contact lists and the plan itself must exist before the incident, stored where the attacker cannot reach or alter them.
- Verify who is on the channel. Membership by named invitation, roll-call checks and code words stop an attacker joining the bridge the way they joined production calls.
- Test it like a backup. Quarterly activation drills reveal expired licences, lost tokens and staff who never installed the app.
Threat assumptions: what the attacker may see
Decide in advance which compromise scenarios trigger OOB use. The broader the compromise, the more independent the channel must be.
| Compromise | What the attacker can observe | ATT&CK | What OOB must avoid |
|---|---|---|---|
| One or more mailboxes | Incident emails, forwarded alerts, legal advice, invitations with dial-in details | T1114 Email Collection; T1114.002 Remote Email Collection | Discussing the response by corporate email; sending invites from compromised calendars |
| Chat and collaboration tenant | War-room channels, files shared there, presence and meeting details | T1213.005 Data from Information Repositories: Messaging Applications | Using the same tenant, even a "private" channel |
| Identity provider or directory | Ability to create accounts, reset MFA, join any SSO-backed app | Account creation and manipulation | Any OOB tool that federates with the compromised identity provider |
| Endpoints of responders | Screen, keystrokes, local files and any app on the device | Collection from the local system | Using managed corporate laptops or phones for OOB |
| Ticketing and documentation | Investigation notes, indicators, containment plans | T1213 Data from Information Repositories | Recording the response in the production ITSM or wiki |
| Internal messaging as an attack channel | Ability to send convincing messages as staff | T1534 Internal Spearphishing | Trusting instructions that arrive on corporate channels during the incident |
CISA's advisory on Scattered Spider (AA23-320A) states that the actors frequently join incident remediation and response calls and teleconferences, likely to see how security teams are hunting them. See the Scattered Spider Threat Actor Profile.
Channel options compared
| Channel | Strengths | Weaknesses | Best use |
|---|---|---|---|
| Separate collaboration tenant (chat, video, files) with its own identity | Familiar tools, file sharing, persistent history for the incident record | Costs a licence and administration; must not federate with production identity | Core responder war room for medium and large organisations |
| End-to-end encrypted messenger with managed groups | Fast to adopt, works on personal or dedicated phones, strong confidentiality | Weak central administration; device loss and backup settings vary; retention for legal hold needs thought | Executive group and first alert; small organisations |
| Dedicated conference bridge with per-incident PIN | Works from any phone; no app | Dial-in details leak easily; hard to verify participants | Large status calls with verified roll call |
| Phone tree and SMS | Universal; CISA's ransomware guide cites phone calls as an OOB method | SMS is not confidential; numbers are often stored only in the directory | Activation and call-out, not detailed discussion |
| IR retainer provider's platform | Already outside your estate; provider supports it | Depends on the provider's onboarding speed and your staff having accounts | When a retainer is invoked; see the Incident Response Retainer Buyer's Guide |
| Satellite phones or radio | Independent of internet and mobile networks | Cost, training, limited users | Physical sites, OT and critical infrastructure where networks may fail |
Decision guidance:
- Small organisations: an end-to-end encrypted messenger group of 10 to 20 named people, a printed call tree, and an agreed fallback conference number is enough, provided accounts are tied to personal or dedicated phones rather than corporate SSO.
- Medium and large organisations: a separate collaboration tenant with its own identity for the response team, an encrypted messenger group for executives, and a conference bridge with roll-call procedures. Large groups should also plan regional or business unit sub-channels.
- Regulated and critical infrastructure: add a channel that survives loss of internet or mobile networks for site operations, and agree how regulators and authorities are contacted outside corporate email.
Designing the OOB environment
Identity
- Separate identity plane: OOB accounts must not authenticate through the production identity provider or directory. Use local accounts in the OOB tool or a separate small identity tenant with its own administrators.
- Phishing-resistant MFA: enrol FIDO2 security keys or passkeys for every OOB account, held by the user, with a spare for administrators. See the Phishing-Resistant MFA Guide.
- Administrators: two or three named OOB administrators who are not production global administrators, plus a sealed break-glass credential. See the Break-Glass Account Guide.
- Personal email for account recovery is a risk: set recovery to hardware keys or a separate admin process, never to a corporate mailbox.
- No integrations: keep bots, connectors and API tokens out of the OOB tenant; each is a credential that could link it back to production.
Devices
- Dedicated or personal devices: issue a small pool of dedicated phones or laptops to core responders, or allow personal devices with the OOB app only, not managed by the production MDM, whose admin console may be compromised.
- Charged, updated and stored: keep pool devices updated monthly with chargers, and a list of who holds which device.
Information to hold offline
Microsoft's incident response planning guidance recommends preparing for email and collaboration impairment, ransomed documentation and unavailable phone numbers, and storing phone numbers, topologies, build documents and restoration procedures immutably offline. NIST SP 800-184 also lists out-of-band communications that do not rely on production systems as a recovery playbook element.
- Contact list with personal mobile numbers for responders, executives, legal counsel, insurers, IR retainer, key suppliers and regulators. Use the Incident Response Contact and Call Tree Template.
- The incident response plan, OOB activation procedure, network diagrams and recovery runbooks, printed and in an encrypted offline store.
- Pre-agreed code words for activation and identity verification.
Data handling and retention
- Retention by design: disappearing messages are attractive for confidentiality but can destroy records you are later obliged to produce. Agree with counsel which OOB channels keep history, for how long, and how a legal hold is applied.
- Export path: make sure the incident log, decisions and shared files can be exported from the OOB environment into your evidence store once production is trusted again, with timestamps preserved.
- Sensitive data minimisation: share indicators, scope and decisions in OOB, but keep raw forensic images, credential material and personal data in the forensic environment, referenced by case number rather than pasted into chat.
- Data location: check where the OOB provider stores messages and files, and whether that is compatible with the regulators and jurisdictions you report to.
Activation and operating rules
- Triggers: activate OOB on any suspected compromise of email, collaboration, identity provider, privileged accounts or the security tooling, and on any ransomware event. When in doubt, activate; it costs little.
- Call out: the incident commander or deputy sends the activation code by phone or messenger to the core group; nobody announces it on corporate channels.
- Roll call: each participant is verified by voice or video against the offline list, and new members are added only by the incident commander after verification. Treat unknown participants as hostile and remove them.
- Move the conversation: all discussion of indicators, scope, containment, legal advice and communications moves to OOB. Corporate channels carry only neutral messages, so as not to tip off the attacker.
- Record: keep the incident log in the OOB environment, with timestamps and decisions, so it can later be preserved for regulators, insurers and litigation. Agree with counsel how privileged advice is labelled. See the Legal Privilege in Incident Response Guide.
- Verify instructions: any instruction that arrives on a corporate channel, or by an unexpected call, is confirmed through OOB before action. Attackers impersonate executives and IT staff during incidents. See the Deepfake CEO Fraud Response Playbook.
- Stand down: return to production channels only when the identity tier and collaboration platform are confirmed clean, and keep the OOB environment for the next incident, rotating any credentials exposed during this one.
Wider stakeholder communications
| Audience | Channel during compromise | Watch out for |
|---|---|---|
| All staff | Pre-registered SMS or mass notification service not tied to the directory; intranet only if trusted | Attackers sending fake "IT instructions"; give staff a way to verify |
| Executives and board | Encrypted messenger group and verified calls | Executive mailboxes are prime targets, as Midnight Blizzard showed |
| Legal counsel and insurer | Phone and OOB tenant guest accounts | Privileged advice sent to compromised mailboxes |
| IR retainer and suppliers | Provider's platform or OOB guest access | Suppliers emailing your compromised addresses |
| Regulators and authorities | Official portals and phone numbers held offline | Reporting deadlines continue whatever your email status; see the Crisis Management Guide |
| Customers | Website status page hosted independently, social accounts with hardware-key MFA | Hijacked social or status accounts used to spread false updates |
Implementation plan
- Scope (week 1 to 2): name the core responder group, executive group and their deputies; agree triggers with legal and the CISO.
- Choose channels (week 2 to 4): select the primary and fallback channels using the comparison above; check data location and retention.
- Build (week 4 to 8): create the separate tenant or groups, local accounts, hardware key enrolment, device pool and offline document set. Disable integrations and external sharing by default.
- Pilot: run a no-notice activation for the core group and measure time to full roll call; fix what failed.
- Roll out: extend to executives, legal and communications, with a 15 minute walkthrough for each person.
- Exercise: include OOB activation in every tabletop. See the Tabletop Exercise Guide.
- Maintain: quarterly membership review against joiners and leavers, monthly device updates, annual review of the channel choice.
Rollback: if the OOB channel itself is suspected compromised (a lost key, an unknown member, a device infection), switch to the fallback channel with a new code word, revoke the affected account and rebuild the group membership from the offline list.
Monitoring and metrics
- Time to assemble: minutes from activation call to verified roll call of the core group in drills (target under 30 minutes).
- Enrolment coverage: percentage of named responders and executives with an active OOB account, enrolled hardware key and device.
- Membership hygiene: number of OOB accounts belonging to leavers or unknown users at each quarterly review (target zero).
- Contact accuracy: percentage of offline contact entries that reached the person on the last drill.
- Drill frequency: number of activation drills per year (at least four) and issues closed from each.
- OOB tenant audit: review sign-in and admin logs of the OOB environment after each drill and incident; alert on new administrators or external sharing.
Common failure patterns
- "Out of band" that shares SSO: the backup chat tool signs in through the compromised identity provider, so the attacker can reset access or join.
- War room in the production tenant: a private Teams or Slack channel created after the incident starts, visible to anyone with tenant admin rights or a compromised member account.
- Dial-in details by email: the bridge number and PIN go out by corporate calendar invite, and the attacker joins the call.
- Contacts only in the directory: nobody has personal numbers once Exchange and the directory are encrypted.
- Never tested: licences expired, executives never installed the app, hardware keys are in a drawer at the office that is closed.
- Legal and executives left on email: the technical team moves to OOB, but counsel, the board and communications staff keep exchanging drafts and advice through mailboxes that are in scope.
Lessons from incidents
- Attackers watch the response: in the Microsoft Midnight Blizzard breach, the actor accessed mailboxes of senior leadership, cybersecurity and legal staff and searched them for information related to Midnight Blizzard itself. Any response coordinated by those mailboxes would have been visible.
- Attackers join the calls: CISA's AA23-320A describes Scattered Spider searching Slack, Teams and Exchange for conversations about the intrusion and joining remediation calls. Verified membership and channels outside the production identity plane are the direct countermeasure.
Framework mapping
| Framework | Reference | Relevance |
|---|---|---|
| NIST SP 800-53 Rev. 5 | IR-4 Incident Handling; IR-8 Incident Response Plan; IR-7 Incident Response Assistance | OOB activation and communication in the IR plan |
| NIST SP 800-53 Rev. 5 | CP-2 Contingency Plan; CP-8 Telecommunications Services | Alternate communications that survive the loss of primary services |
| NIST SP 800-53 Rev. 5 | IA-2 Identification and Authentication; SC-8 Transmission Confidentiality and Integrity | Separate, strongly authenticated, encrypted channels |
| NIST CSF 2.0 | RS.MA-01; RS.CO-02; RS.CO-03; PR.AA-03 | Coordinated response, stakeholder notification and authenticated users |
| NIST SP 800-61 Rev. 3 and SP 800-184 | Incident response aligned to CSF 2.0; OOB communications in recovery playbooks | Planning guidance |
| CIS Controls v8.1 | Control 17 Incident Response Management | Contact information and communication mechanisms for incidents |
| ISO/IEC 27001:2022 | A.5.24 Information security incident management planning and preparation; A.5.26 Response to information security incidents | Prepared response procedures |
Practitioner checklist
- Name the core responder and executive OOB groups with deputies, and agree activation triggers.
- Choose a primary and a fallback channel that do not use the production identity provider, MDM or devices.
- Create OOB accounts with hardware security keys and two named OOB administrators who are not production admins.
- Hold personal contact details, the IR plan, diagrams and runbooks offline and printed.
- Agree code words and a roll-call verification procedure, and add members only through the incident commander.
- Keep the incident log in the OOB environment and agree privilege labelling with counsel.
- Run a no-notice activation drill every quarter and measure time to verified roll call.
- Review OOB membership against leavers quarterly and audit OOB admin logs after each drill.
Standards and references
- CISA: AA23-320A Scattered Spider
- CISA: #StopRansomware Guide
- NIST: SP 800-61 Rev. 3 Incident Response Recommendations and Considerations for Cybersecurity Risk Management
- NIST: SP 800-184 Guide for Cybersecurity Event Recovery
- Microsoft Learn: Incident response planning
- NIST: SP 800-53 Rev. 5
- MITRE ATT&CK: Email Collection
Related NHI Mgmt Group resources: Crisis Management Guide · Out-of-Band Communications Plan Template · Incident Response Contact and Call Tree Template · Identity Incident Response Playbook