Firewalls, VPN gateways, routers and load balancers sit on the internet edge, run vendor firmware you cannot scan like a server, and are now among the most exploited systems in the CISA Known Exploited Vulnerabilities catalogue. Vulnerability management processes built for Windows and Linux fleets do not fit them well: there is no agent, updates mean a reboot of something the whole business depends on, advisories arrive on each vendor's own schedule, and end-of-support dates quietly end the supply of fixes. This guide covers how to inventory network devices, track vendor advisories, prioritise, plan maintenance windows, upgrade firmware safely, and handle devices that cannot be patched. It is for network, infrastructure and vulnerability management teams. General patching process is covered in the Patch Management Guide; this page covers what is different for network and edge kit.
Key takeaways
- Edge devices need an emergency lane. Mass exploitation of edge flaws can start within a day of disclosure, as with Ivanti Connect Secure in January 2024, so a quarterly network maintenance cycle is not an adequate response to a KEV entry.
- Inventory must include software train and support dates. Without model, exact firmware version, feature set and end-of-support date for every device, advisory matching is guesswork.
- Use the vendor's own advisory tooling. Network vendors publish fixed-release information in their own formats and tools (for example the Cisco Software Checker and CSAF downloads); generic CVE feeds lag and miss train-specific detail.
- Patching does not undo a compromise. If a device was exposed while vulnerable, check integrity, rotate the secrets stored on it and in its configuration, and consider rebuild before trusting it.
- Management plane exposure multiplies risk. Removing internet access to management interfaces removes a large share of edge risk before any firmware is touched.
Why network devices are different
| Factor | Servers and endpoints | Network and edge devices | What to do about it |
|---|---|---|---|
| Visibility | Agents report installed packages | No agent; version from SNMP, API, CLI or management platform | Collect versions through the vendor management platform or configuration backup tooling, and reconcile with scanner fingerprints |
| Update unit | Individual packages | Whole firmware image or software train | Choose target releases deliberately; a fix may require moving trains |
| Downtime | Rolling restarts behind load balancers | Reboot interrupts traffic unless devices are clustered | Design high-availability pairs so one member can be upgraded at a time |
| Advisory cadence | Monthly vendor cycles | Vendor-specific schedules plus frequent out-of-band releases for exploited flaws | Subscribe to each vendor's PSIRT channel and API |
| Lifecycle | OS support lifecycles well tracked | End-of-sale and last-day-of-support dates per model and train | Track support dates in the inventory and budget replacement ahead |
| Forensics | EDR telemetry and disk images | Limited logs, appliance shells closed to customers | Use vendor integrity tools and central logging; see the Edge Device Compromise Investigation Playbook |
The local copy of the CISA KEV catalogue (version of 21 October 2026) shows how much of the exploited-vulnerability flow comes from network vendors: in 2026 to that date, 19 entries were added for Cisco, 9 for Fortinet, 6 for Ivanti and 6 for Citrix. See the Edge Device Exploitation Guide for the threat picture.
Inventory and advisory intelligence
What the inventory must hold
- Identity: hostname, model, serial number, location, owner and business service supported.
- Software: exact running version and software train, installed feature licences and enabled features (VPN portal, web management, Smart Install, SNMP, specific routing protocols), because many advisories apply only when a feature is on.
- Exposure: whether any interface, especially management, VPN or web portal, is reachable from the internet.
- Lifecycle: end-of-sale, end-of-software-maintenance and last-day-of-support dates.
- Redundancy: cluster or pair membership, which determines whether it can be upgraded without an outage.
Advisory sources by type
| Source | What it gives you | How to use it |
|---|---|---|
| Vendor PSIRT advisories and APIs | Affected and fixed releases per train, workarounds, exploitation status | Automate ingestion (for example Cisco's openVuln API, RSS or CSAF files) and match against inventory |
| Vendor fixed-release tools | The minimum release that fixes all known advisories for your version | Use them to choose a target release, not just to fix one CVE |
| CISA KEV catalogue | Confirmed exploitation and required action | Any KEV match on an edge device goes straight to the emergency lane; see the CISA KEV Catalogue Guide |
| Joint government advisories | Indicators, integrity checks, hardening guidance | Feed into hunting and configuration baselines |
| Authenticated or fingerprint scanning | Independent confirmation of version and exposure | Validate the inventory; catch devices missing from it |
Know each vendor's rhythm
Cisco is a useful example of why vendor schedules matter. Its published policy states that security advisories are generally disclosed at 1600 GMT on Wednesdays, with bundled publications twice a year: FXOS and NX-OS on the fourth Wednesday of February and August, IOS XR on the second Wednesday of March and September, and IOS and IOS XE on the fourth Wednesday of March and September, plus out-of-cycle publication when there is active exploitation. Cisco's Security Impact Rating follows CVSS base score bands (Critical 9.0 to 10.0, High 7.0 to 8.9, Medium 4.0 to 6.9). Build each major vendor's schedule into the change calendar so assessment capacity is ready on publication day.
Prioritisation for network kit
Rank network device findings on exposure and exploitation first, CVSS second. A medium-rated flaw in an internet-facing VPN portal that is in the KEV catalogue outranks a critical flaw in a feature that is disabled on every device.
| Tier | Criteria | Target response |
|---|---|---|
| Emergency | In KEV or vendor-confirmed exploitation, on an internet-reachable interface or feature that is enabled | Apply vendor mitigation or restrict access within hours; upgrade in an emergency window within days |
| Urgent | Critical or high, pre-authentication, enabled feature, reachable from untrusted networks, not yet exploited | Next expedited window, typically within two weeks |
| Planned | Requires authentication or internal adjacency, or affects a feature that is disabled | Next scheduled maintenance cycle |
| Lifecycle | Device at or past last day of support | Replacement project with isolation in the meantime |
Use the SSVC Guide and the EPSS Guide for the decision model, and set the targets in your remediation SLAs.
Maintenance windows and safe firmware upgrades
NIST SP 800-40 Rev. 4 recommends grouping assets into maintenance groups with similar characteristics and defined responses for routine and emergency scenarios, and staged rollouts that start with canary assets. For network devices that translates into:
- Define maintenance groups: for example internet edge firewalls and VPNs, data centre core, campus access, branch routers, and out-of-band management. Each group has its own routine window and an emergency procedure.
- Pick a target release, not just a fix: use the vendor's fixed-release tool to find a release that closes all open advisories for the train, and prefer the vendor-recommended release. The joint CISA, NSA, FBI and Five Eyes guidance on communications infrastructure advises running the vendor-recommended OS version and testing patches before and after deployment.
- Verify the image: download from the vendor portal only, check the published hash or signature, and store approved images in an internal repository.
- Back up and record state: save the running configuration, licence state, routing tables and session counts before the change, so you can compare after.
- Upgrade canaries first: one lab device or one member of a non-critical pair, then monitor for a defined period.
- Roll through pairs: upgrade the standby member, fail over, verify, then upgrade the other. For single devices, schedule the outage with the service owner.
- Verify: confirm the running version, compare routing, VPN and session state with the baseline, and run an authorised exposure check for the CVE.
- Rollback plan: keep the previous image on the device where the platform supports dual images, and document the downgrade path. Some upgrades change configuration formats, so test downgrade in the lab.
Emergency lane
- Hours, not days: apply the vendor's published mitigation or remove internet access to the affected interface first, then patch.
- Pre-approved change: an emergency change class for KEV-listed edge flaws avoids waiting for the next change advisory board. See the Emergency Out-of-Band Patching Playbook.
- Integrity check and hunt: run the vendor's integrity checking tool where one exists and review logs for the advisory's indicators. Treat a clean integrity check as one data point, not proof: in the Ivanti Connect Secure exploitation, CISA warned that the vendor's integrity checker failed to detect compromise and that credentials stored on affected appliances should be presumed compromised.
Reducing exposure while you wait
CISA Binding Operational Directive 23-02, issued on 13 June 2023 for US federal civilian agencies, is a useful model for any organisation. It covers routers, switches, firewalls, VPN concentrators, proxies, load balancers and out-of-band server management interfaces such as iLO and iDRAC, and requires that management interfaces exposed to the internet over protocols such as HTTP, HTTPS, SSH, SNMP, Telnet, TFTP and SMB be removed from the internet within 14 days, or protected by a separate zero trust policy enforcement point.
- Management plane: move management to an out-of-band network reached only from administrative workstations. See the Network Device Management Plane Hardening Guide.
- Disable unused features: many network advisories apply only to a specific feature, such as a web UI, a VPN portal or Smart Install. Turning it off removes the exposure.
- Virtual patches: IPS signatures and access lists in front of a device buy time but rarely cover appliance flaws fully. See the Virtual Patching Guide.
- End-of-support devices: Cisco's policy, for example, states that after the last day of support its PSIRT no longer analyses, fixes or discloses vulnerabilities, and issues no CVEs for those products. Silence in the advisory feed for an unsupported device means no fixes, not no flaws. Isolate it and replace it; see the Edge Device End of Life Replacement Playbook.
Small estates versus large estates
- Small organisations (a handful of firewalls and switches, often managed by a provider): the priorities are knowing exactly which models and versions you run, receiving the vendor's advisory emails directly rather than relying on the provider, keeping management off the internet, and writing the emergency patch expectation into the managed service contract with a response time in hours. Replacing a single unsupported firewall is usually cheaper than the effort of isolating it.
- Large organisations (thousands of devices across many vendors): automate advisory ingestion through vendor APIs and CSAF, standardise on a small number of approved releases per platform, and run upgrades through a network automation pipeline that takes the backup, verifies the image, upgrades, checks state and rolls back on failure. Track each platform's "approved release" as a configuration baseline, so drift from it shows up as a finding.
- Managed and cloud-delivered devices: where the provider patches the edge service for you, the remaining work is contractual and verification: confirm the provider's patch timelines for exploited flaws, and check its advisories for actions customers must take themselves, such as rotating credentials.
Where identity fits
Network devices are full of non-human credentials: local administrator accounts, SNMP community strings, TACACS+ and RADIUS shared keys, VPN pre-shared keys, API tokens for management platforms and certificates. In the Salt Typhoon telecom intrusions, one entry point was a roughly seven-year-old Cisco Smart Install flaw (CVE-2018-0171), but most access used valid credentials, and stolen configurations exposed SNMP strings and weakly encrypted local passwords that let the attackers move between devices. Vulnerability management for network kit must therefore include configuration findings: weak password storage, default or shared SNMP strings, and AAA keys that have never been rotated. Rotate these after any suspected compromise and protect configuration backups as secrets. See the Secrets Management Guide.
Detection and measurement
Send device logs and AAA logs to a central platform over encrypted transport, keep copies that cannot be altered, and alert on configuration changes made outside change management, new local accounts or SSH keys, changed AAA servers, and unexpected reboots or version changes. The joint communications infrastructure guidance recommends treating a central configuration store as the source of truth and alerting on drift. See the Network Device Log Detection Guide.
- Percentage of network devices with model, version and support date in the inventory.
- Time from vendor advisory publication to completed applicability assessment.
- Time to mitigate KEV-listed flaws on internet-facing devices, measured in hours.
- Percentage of devices on a release with no open critical or high advisories.
- Number of devices past last day of support, and how long each has been there.
- Number of management interfaces reachable from the internet (target zero).
Common failure patterns
- Old flaws left open: the Salt Typhoon case shows a 2018 vulnerability still exploitable in 2025. Long-lived devices accumulate unpatched advisories when upgrades are deferred for fear of outages.
- Patch without hunt: teams upgrade a VPN gateway after a KEV listing but never check whether it was already compromised, leaving persistence and stolen credentials in place.
- Inventory without versions: devices are listed by model only, so applicability checks for train-specific advisories cannot be automated.
- Single devices with no redundancy: because every upgrade means an outage, upgrades are postponed indefinitely.
- Unsupported kit kept "because it works": devices past last day of support receive no fixes and no new CVEs, so scanners and dashboards show them as clean.
- Management on the internet: web and SSH management interfaces left exposed turn every management-plane advisory into an emergency.
Framework mapping
| Framework | Reference | Relevance |
|---|---|---|
| NIST SP 800-40 Rev. 4 | Maintenance groups, routine and emergency patching | Planning model for network maintenance windows |
| NIST CSF 2.0 | ID.AM-08, ID.RA-01, PR.PS-02 | Lifecycle management; vulnerability identification; software maintained and replaced commensurate with risk |
| NIST SP 800-53 Rev. 5 | CM-8, RA-5, SI-2, SA-22, AC-17 | Inventory; scanning; flaw remediation; unsupported components; remote access |
| CIS Controls v8.1 | Safeguard 12.1, Control 7 | Ensure network infrastructure is up to date; continuous vulnerability management |
| ISO/IEC 27001:2022 Annex A | A.8.8, A.8.9 | Management of technical vulnerabilities; configuration management |
| MITRE ATT&CK v20 | T1190, T1133, T1556.004, T1601, T1542.001 | Exploit Public-Facing Application; External Remote Services; Network Device Authentication; Modify System Image; System Firmware |
Practitioner checklist
- Inventory every network device with model, exact version, enabled features, exposure and support dates.
- Subscribe to each vendor's PSIRT feed or API and match advisories to inventory automatically.
- Define maintenance groups with routine windows and a pre-approved emergency change for KEV-listed edge flaws.
- Build high-availability pairs for edge and core devices so upgrades do not need outages.
- Verify image hashes, back up configuration and state, upgrade canaries first and keep a tested rollback.
- Remove management interfaces from the internet and disable unused features.
- After any exposure window, run integrity checks, hunt, rotate device credentials and AAA keys, and consider rebuild.
- Track devices past last day of support and isolate them until replacement.
- Alert on configuration drift, version changes and new local accounts.
Standards and references
- NIST: SP 800-40 Rev. 4 Guide to Enterprise Patch Management Planning
- CISA: Binding Operational Directive 23-02, Mitigating the Risk from Internet-Exposed Management Interfaces
- CISA, NSA, FBI and partners: Enhanced Visibility and Hardening Guidance for Communications Infrastructure
- CISA: Known Exploited Vulnerabilities Catalog
- Cisco: Security Vulnerability Policy
- NIST: SP 800-53 Rev. 5
- MITRE ATT&CK: Modify System Image
Related NHI Mgmt Group resources: Patch Management Guide · Edge Device Security Guide · Vulnerability Management Guide · Firmware and Boot Security Guide