A public warning system turns one authenticated message into a siren on millions of phones, televisions and radios, so a stolen originator credential, a confusing software screen or a missing second check can cause public panic as surely as an attacker can. This guide covers the security of alert origination: who is allowed to send, how originator credentials and digital certificates are protected, how two-person release and test separation prevent false alerts, and how to retract one fast when it happens anyway. It uses the US Integrated Public Alert and Warning System (IPAWS), the Emergency Alert System (EAS) and Wireless Emergency Alerts (WEA) as the main worked example, with notes for EU public warning under the European Electronic Communications Code. It is for emergency management agencies, alerting authority administrators, 911 and public safety IT teams, and broadcasters operating EAS equipment. Choosing a notification platform is covered in the Mass Notification Systems Buyer's Guide, and 911 centre security in the Emergency Services and NG911 Security Guide; this page does not repeat either.
Key takeaways
- Origination is the crown jewel. Once an alerting authority's software holds a valid IPAWS certificate and permissions, downstream channels trust what it sends, so the controls that matter most sit at the originator, not in the distribution network.
- False alerts are a security failure mode. The FCC's investigation of the January 2018 Hawaii false missile alert attributed it to a combination of human error and inadequate safeguards, and the correction took 38 minutes.
- Two credentialed people for high-impact alerts and all tests. The FCC recommended more than one credentialed person verify ballistic missile alerts, other high-impact alerts and all tests, with live and test environments separated in the software.
- Pre-built retractions save minutes. CAP supports Cancel and Update message types; have correction templates authorised in advance so a retraction is a release decision, not a drafting exercise.
- Regulation is tightening. In June 2026 the FCC adopted EAS rules on passwords, prompt security patching and firewalls, and opened further proposals including authentication of all alerts before transmission.
How alert origination works and where it can fail
In the US, an alerting authority (a state, local, tribal or territorial agency) writes an alert in IPAWS-compatible software, which formats it as a Common Alerting Protocol (CAP) message and sends it to IPAWS-OPEN. IPAWS checks the sender's credentials and permissions and relays the alert to EAS (broadcast and cable), WEA (mobile phones), NOAA Weather Radio and internet services. Broadcasters' EAS encoder and decoder units also receive and relay alerts. Each step is a trust boundary.
| Stage | What it trusts | Main failure modes | Primary control |
|---|---|---|---|
| Originator workstation and user | The person logged in to alert software | Stolen or shared credentials, wrong template chosen, drill confused with live | Named accounts, phishing-resistant MFA, two-person release, distinct live and test modes |
| Alert origination software | The COG digital certificate configured in it | Certificate copied or exported, software exposed to the internet, weak admin roles | Certificate custody, admin role separation, hosting hardening, vendor security updates |
| IPAWS-OPEN | Certificate and permissions set at onboarding | Over-broad permissions (alert types or geography beyond need) | Least privilege in the public alerting permissions application, periodic review |
| EAS participants (broadcast, cable) | Alerts received from IPAWS or monitored sources | Internet-exposed encoder/decoder management interfaces, default passwords, unpatched firmware | FCC 2026 password, patching and firewall requirements; management network isolation |
| Public reception | The channel itself | Rumour and spoofed social media posts during and after an incident | Pre-planned official social media statements, consistent retraction channels |
Originator onboarding and least privilege
FEMA's sign-up process sets the identity foundations. An agency consults its state IPAWS representative, has staff complete IS-247 (the alert originator course; IS-251 for alerting administrators is optional), buys IPAWS-compatible software tested in the IPAWS-OPEN test environment, applies for a Memorandum of Agreement (MOA) to become a Collaborative Operating Group (COG), receives a digital certificate with the executed MOA, then applies for public alerting permissions that define alert types and geographic area, signed by a designated state official or tribal leadership. Use each step as a control point:
- Permissions: request only the event codes and geography the agency is responsible for. A county should not hold statewide permissions "just in case"; mutual aid can be handled by the state or a neighbouring COG.
- Training as an access gate: tie account creation in the alert software to proof of IS-247 completion, and remove accounts when staff leave the role, not just the agency.
- Roles: separate the administrator who manages users, templates and certificate configuration from the originators who draft and release alerts. No originator should be able to edit the templates they send from.
- Software selection: security requirements for the origination platform (SSO, MFA, audit export, role model) belong in procurement. See the Mass Notification Systems Buyer's Guide.
Protecting originator credentials and the COG certificate
The COG certificate is a machine credential that represents the whole agency to IPAWS. Treat it as a high-value non-human identity: anyone who holds it and the software configuration can send alerts the network will treat as authentic.
| Asset | Risk | Control |
|---|---|---|
| COG digital certificate | Exported to laptops, emailed to vendors, left in old installs | Install only on the production origination platform; record every location; restrict export; request replacement through IPAWS if exposure is suspected |
| Originator accounts | Shared duty-desk logins, password reuse, phishing | Named accounts, phishing-resistant MFA, no shared logins even for 24/7 desks |
| Administrator accounts | Template tampering, user creation, permission changes | Separate privileged accounts, MFA, change logging reviewed by a second person |
| Vendor support access | Persistent remote access into hosted or on-premises alert software | Time-bound, approved sessions with logging; contract terms on support staff vetting |
| Integration and API credentials | Feeds from weather, CAD or sensors that can auto-generate alerts | Individual service identities, scoped to draft-only where possible, rotated on staff or vendor change |
Because the certificate is effectively a long-lived machine secret, apply the same custody discipline as any other signing credential. See the Break-Glass Account Guide for handling the emergency accounts that 24/7 duty desks inevitably need.
Two-person release and false-alert prevention
The FCC Public Safety and Homeland Security Bureau's final report on the Hawaii false alert found that a warning officer sent a live ballistic missile alert believing an attack was real, when an internal test was intended. Its recommendations to emergency management agencies included testing in a closed environment such as the FEMA IPAWS Test Lab, requiring more than one credentialed person to verify high-impact alerts and all tests, asking software providers to separate live and test environments and confirm alert type before sending, limiting who can create or modify drill messages, supervisor approval of internal tests, documented false-alert procedures and planned social media use.
Design rules for the release workflow
- Classify alert templates by impact: for example tier 1 (attack warning, nuclear or evacuation of large areas), tier 2 (shelter in place, civil emergency), tier 3 (local advisories, amber alerts as your policy allows). Define the release rule per tier.
- Two-person release for tier 1 and 2 and every test: one person drafts, a second credentialed person reviews the text, area and status and releases. The second person authenticates separately; a shared screen click does not count. This is NIST SP 800-53 AC-3(2) Dual Authorization applied to alerting. See the Dual control and four-eyes approval for privileged and agent actions guide.
- Make test and live visually and technically different: separate environments or accounts, different screen colours or banners, and CAP status set correctly (Actual, Exercise, System, Test or Draft). Live mode should require a deliberate extra step.
- Show the operator what will happen: the confirmation screen states alert type, channels (WEA, EAS, both), geography and estimated reach in plain words, not a generic "Are you sure?".
- Script drills so they cannot be mistaken for reality: drill messages should not contain live alert wording, and only supervisors should create or alter drill scripts.
- Keep a solo path only for genuine time-critical hazards: where policy allows one person to send (for example a tornado warning with minutes of lead time), limit it to specific templates and review every solo release afterwards.
Retraction and correction
CAP 1.2, an OASIS Standard, defines Update and Cancel message types that reference earlier messages, so a correction can target the false alert directly. The Hawaii correction took 38 minutes. Plan so yours takes minutes:
- Pre-authorised correction templates: for every tier 1 and tier 2 template, hold a matching "false alert, no threat" Cancel or Update text approved in advance, available to the same channels and geography.
- Retraction authority: name who can release a correction alone. A faster single-person path for retractions is a reasonable trade-off, because the harm of delay is greater than the harm of a mistaken correction; log and review every use.
- Parallel channels: pre-drafted social media posts, a media line and a script for 911 call takers, since call volumes spike. The FCC also found Hawaii's phone lines were overwhelmed and recommended redundant communications such as GETS and WPS.
- Reporting: under 47 CFR 11.45(b), an EAS Participant that discovers it has sent a false alert must email the FCC Operations Center within 24 hours; government entities are encouraged to do the same. Add this to the false-alert runbook with the address and the facts to include.
Securing EAS equipment and origination infrastructure
On 25 June 2026 the FCC adopted a Report and Order requiring EAS participants to adopt stronger password practices, promptly test and install manufacturer security patches, and protect alerting equipment with network firewalls or comparable measures. A Further Notice of Proposed Rulemaking seeks comment on authentication for all alerts before transmission, a universal alert identifier and other changes. Compliance dates and final text should be confirmed in the FCC record before relying on them; an earlier 2022 proposal (FCC 22-82) had floated 72-hour reporting of unauthorised access to EAS equipment and annual cybersecurity plan certification.
- No internet-facing management: encoder/decoder and origination server management interfaces sit on an isolated management network reachable through an authenticated jump path, never directly from the internet.
- Change default credentials on day one: include every local account on EAS units and appliances, and store them in a vault.
- Patch promptly: subscribe to manufacturer security notices and track firmware levels in the asset register. See the ICS Advisory Triage and OT Vulnerability Management Guide for an approach to embedded equipment.
- Hosted origination platforms: require SSO, MFA, audit log export and documented tenant isolation; restrict admin consoles by network location where the platform supports it.
- Resilience: keep a second origination path (backup site, second workstation or state-level fallback) and exercise it, because an agency that cannot send a real warning has also failed.
EU and other jurisdictions
Article 110 of the European Electronic Communications Code (Directive (EU) 2018/1972) required Member States, by 21 June 2022, to ensure that where public warning systems exist, mobile number-based interpersonal communications providers transmit warnings to affected end-users. Member States may instead use other public electronic communications services or a mobile app, if as effective in coverage and capacity, with regard to BEREC guidelines. The origination controls in this guide apply equally to cell broadcast and location-based SMS systems operated under national schemes: the authority that originates holds the credential, so the same two-person release, test separation and retraction planning apply. National operating rules differ, so confirm the scheme operator's own security requirements.
Detection and measurement
Log every draft, approval, release, cancel and update with user, time, template, geography and CAP status. Review weekly, and alert in real time on:
- A release of a tier 1 or tier 2 template, or any Actual-status message from an account that normally sends tests.
- Logins to origination software outside duty patterns or from new locations; administrator changes to templates, users or permissions.
- Certificate configuration changes or new installations of origination software.
- EAS units receiving alerts with unexpected originator codes or outside expected geography (for EAS participants).
Metrics
- Percentage of tier 1 and 2 releases and tests with two-person approval (target 100%).
- Time from false-alert discovery to correction release, measured in exercises (target: minutes, set per tier).
- Originator accounts with MFA and current IS-247 training (target 100%).
- Number of locations holding the COG certificate versus the documented list.
- Days to apply EAS and origination platform security patches.
- Successful closed-environment test releases per quarter.
Framework mapping
| Control area | NIST SP 800-53 Rev. 5 | NIST CSF 2.0 | Threat addressed (ATT&CK) |
|---|---|---|---|
| Two-person release | AC-3(2) Dual Authorization, AC-5 Separation of Duties | PR.AA-05 | Insider error or misuse |
| Originator authentication | IA-2(1), IA-5 | PR.AA-01 | T1078 Valid Accounts, T1110 Brute Force |
| Internet-facing origination and EAS interfaces | SC-7 Boundary Protection | PR.IR-01 | T1190 Exploit Public-Facing Application, T1133 External Remote Services |
| Template and configuration control | CM-2 | PR.PS-01 | Unauthorised alert content |
| Audit and review | AU-6 | DE.CM-03 | Misuse detection |
| Exercises and retraction | CP-4 Contingency Plan Testing | RS.CO-02 | Delayed correction |
Common failure patterns
- Same screen for test and live: drill and real alerts look identical in the software, so the only safeguard is the operator's attention.
- Shared duty-desk login: a 24/7 desk uses one account, which removes accountability and makes two-person release meaningless.
- Certificate sprawl: the COG certificate sits on old workstations, a vendor's support laptop and a training install.
- No correction ready: the retraction has to be written, approved and permissioned during the incident.
- Exposed equipment: EAS units or origination servers with web management on the internet and default passwords.
- Over-broad permissions: an agency can send event types or areas it never uses, widening the impact of a compromised account.
Practitioner checklist
- Inventory every originator, administrator, integration account and COG certificate location.
- Enforce named accounts with phishing-resistant MFA in the origination software.
- Reduce IPAWS public alerting permissions to the event codes and geography you need.
- Classify templates by impact and require two-person release for high-impact alerts and all tests.
- Separate test and live environments and use correct CAP status values.
- Pre-approve Cancel and Update correction templates for every high-impact template.
- Put the 47 CFR 11.45(b) 24-hour false alert report in the runbook.
- Remove internet exposure of EAS and origination management interfaces; change defaults; patch promptly.
- Exercise a false alert and retraction at least annually in a closed test environment.
- Track the FCC June 2026 rules and further proposals and update controls as compliance dates are set.
Standards and references
- FEMA: Sign Up for IPAWS
- FCC: Report and Recommendations, Hawaii Emergency Management Agency January 13, 2018 False Alert
- FCC: Testimony on the Hawaii False Emergency Alert
- OASIS: Common Alerting Protocol Version 1.2
- US Code of Federal Regulations: 47 CFR 11.45 Prohibition of false or deceptive EAS transmissions
- FCC: Proposals to Strengthen EAS and WEA Cybersecurity (FCC 22-82)
- EUR-Lex: Directive (EU) 2018/1972 European Electronic Communications Code
- Radio Online: FCC Adopts EAS Cybersecurity Modernization Rules
Related NHI Mgmt Group resources: Emergency Services and NG911 Security Guide, Mass Notification Systems Buyer's Guide, Dual control and four-eyes approval for privileged and agent actions, Public Sector Identity Security Guide.