Join our Newsletter — 33% off our NHI Course
Home› Guides› Edge Device Exploitation by Threat Actors Guide: Attack…
Guide Attack Techniques & Malware

Edge Device Exploitation by Threat Actors Guide: Attack Patterns, Detection and Response

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 27 October 2026 13 min read
On this page

VPN gateways, firewalls and remote access appliances have become the most reliable way into an organisation, because a single unauthenticated flaw on an internet-facing box hands an attacker a privileged foothold that endpoint tools cannot see. This guide explains the exploitation pattern that state espionage groups and ransomware affiliates now share: find a zero-day or freshly disclosed bug, compromise the appliance, harvest the credentials and sessions it holds, then persist below the level that patching reaches. It covers what the CISA Known Exploited Vulnerabilities (KEV) catalogue and incident data show, how to detect a compromised device, and how to respond without leaving the attacker inside. It is for SOC analysts, network engineers and vulnerability managers. Hardening baselines are in the Edge Device Security Guide and are not repeated here.

Key takeaways

  • Exploitation is now the leading way in. Mandiant M-Trends 2025 found exploits were the initial infection vector in 33% of its 2024 investigations, and the 2026 Verizon DBIR says 31% of breaches now start with a software vulnerability.
  • Security and networking products are the zero-day target of choice. Google Threat Intelligence Group counted 20 of 33 enterprise zero-days exploited in 2024 in security and networking products, and all five zero-days it attributed to PRC-backed groups hit that category.
  • The appliance is a credential store. Ivanti, Citrix and Cisco compromises exposed VPN user passwords, service account passwords, API keys, certificates and live session cookies, so patching alone does not close the incident.
  • Persistence can survive the patch and the reset. CISA has warned of rootkit-level persistence on Ivanti gateways that survives a factory reset, Fortinet found a symlink that kept read-only access after upgrading, and Cisco ASA campaigns modified ROM.
  • Response is a credential event, not a firmware event. Assume every secret on or passing through the device is exposed: rotate, revoke sessions and Kerberos tickets, and rebuild from a known good image.

How the attack works

The pattern varies by product but follows a recognisable sequence. The technique IDs are MITRE ATT&CK v20.

  1. Reconnaissance at internet scale: attackers scan for exposed management interfaces and VPN portals by product banner and version. After a public advisory, mass exploitation by other actors often follows within days, as with Ivanti Connect Secure in January 2024.
  2. Initial exploitation: an unauthenticated flaw in the web or VPN service, often an authentication bypass chained with command injection, gives code execution on the appliance (T1190 Exploit Public-Facing Application). Ivanti CVE-2023-46805 with CVE-2024-21887 and Palo Alto CVE-2024-3400 in GlobalProtect are typical.
  3. Establish persistence on the device: web shells in appliance web components (T1505.003 Web Shell), patched system binaries (T1601.001 Patch System Image), or boot-level implants (T1542.004 ROMMONkit). CISA ED 25-03 describes an actor manipulating Cisco ASA ROM to persist through reboots and upgrades.
  4. Harvest credentials: the login page is modified to capture plaintext passwords (T1556.004 Network Device Authentication), configuration files with stored secrets are dumped (T1552.001 Credentials In Files), or traffic is captured (T1040 Network Sniffing). Memory disclosure bugs such as CitrixBleed leak session cookies directly.
  5. Pivot inward with valid accounts: stolen domain or service account credentials and hijacked sessions are replayed through the VPN (T1078 Valid Accounts, T1133 External Remote Services). From here the intrusion looks like ordinary user activity.
  6. Cover tracks and impair defences: logs on the appliance are cleared or redirected (T1070 Indicator Removal), and some actors modify integrity checking tools (T1685 Disable or Modify Tools). Compromised devices may also be enrolled as relays for other operations (T1584.008 Network Devices).

GTIG notes that these appliances provide broad network access while EDR tools often cannot monitor them, giving the attacker privilege and invisibility at once.

Where it shows up

The CISA KEV catalogue is the best public record of which edge flaws are exploited. The entries below are drawn from the KEV dataset (catalogue version 2026.10.21). "Known" in the ransomware column means CISA has evidence of use in ransomware campaigns; "Unknown" means no confirmed link, not that it is safe.

CVEProductAdded to KEVRansomware useWhy it matters
CVE-2018-13379Fortinet FortiOS SSL VPN2021-11-03KnownPath traversal leaking VPN session files; still exploited years after patch
CVE-2022-42475Fortinet FortiOS2022-12-13KnownHeap overflow in SSL-VPN, used for pre-auth code execution
CVE-2023-3519Citrix NetScaler ADC and Gateway2023-07-19KnownCode injection, web shells planted on gateways
CVE-2023-4966 (CitrixBleed)Citrix NetScaler ADC and Gateway2023-10-18KnownSession cookies leaked from memory, MFA bypassed
CVE-2023-20198Cisco IOS XE Web UI2023-10-16UnknownPrivilege escalation creating local admin accounts on routers and switches
CVE-2023-46805 and CVE-2024-21887Ivanti Connect Secure and Policy Secure2024-01-10KnownChained auth bypass and command injection; credential harvesting at scale
CVE-2024-3400Palo Alto Networks PAN-OS GlobalProtect2024-04-12KnownUnauthenticated command injection, CVSS-B 10.0
CVE-2024-40766SonicWall SonicOS2024-09-09KnownImproper access control, linked to ransomware intrusions via SSL VPN
CVE-2025-0282Ivanti Connect Secure, Policy Secure, ZTA Gateways2025-01-08KnownStack overflow exploited as a zero-day
CVE-2025-5777 (CitrixBleed 2)Citrix NetScaler ADC and Gateway2025-07-10KnownOut-of-bounds read leaking session tokens
CVE-2025-20333 and CVE-2025-20362Cisco Secure Firewall ASA and FTD2025-09-25UnknownRCE and missing authorisation; subject of CISA ED 25-03

A keyword filter that NHI Mgmt Group ran over the same KEV dataset (VPN, gateway, firewall and named appliance operating systems) returned 168 entries, of which 52 are marked Known for ransomware use. The filter is approximate, but every year since 2022 has added roughly 20 or more such entries from a small group of vendors.

Who uses it: the pattern is shared across motives. China-nexus espionage groups dominate zero-day use against appliances (GTIG attributed five 2024 zero-days to PRC-backed groups, all in security and networking products), and the Salt Typhoon Threat Actor Profile, Volt Typhoon Threat Actor Profile and UNC3886 Threat Actor Profile Guide show long-term device persistence. Ransomware affiliates move fast on n-day bugs once proof-of-concept code is public; see the Akira Threat Actor Profile and LockBit Threat Actor Profile. Compromised devices also end up in relay networks, covered in the Operational Relay Box Networks Guide.

Impact

  • Credential exposure beyond the device: in the Ivanti Connect Secure exploitation of 2024, a credential harvester captured plaintext VPN logins, and configuration dumps exposed service account passwords used for authentication server binding, API keys and certificates.
  • MFA bypass through sessions: the CitrixBleed 2023 compromises showed that leaked session cookies remained valid after patching until sessions were explicitly killed. See the Session Hijacking and Cookie Theft Guide.
  • Network control: on routers and firewalls attackers change AAA and TACACS+ server addresses, add local accounts and SSH keys, and alter ACLs, as in the Salt Typhoon telecom intrusions, where access persisted for years.
  • Ransomware deployment: VPN access leads directly to domain compromise and encryption; KEV marks most of the high profile edge CVEs above as Known for ransomware use.
  • Long dwell time: appliances lack EDR, so discovery often comes from outside. M-Trends 2025 reports a 26 day median dwell time when an external entity notified the victim, against 10 days for internal discovery.

Where identity fits

An edge appliance holds more machine credentials than most servers: LDAP or Active Directory bind accounts, RADIUS and TACACS+ shared secrets, SAML signing certificates, API keys for management platforms, and SNMP community strings. These are non-human identities, usually long-lived and rarely rotated. CISA ED 24-01 supplemental direction V2 told agencies to reset the admin enable password, stored API keys and local and service account passwords, to reset on-premises domain account passwords twice, revoke Kerberos tickets and revoke cloud tokens. Treat that list as the template for any edge compromise. Give appliances dedicated, least-privilege bind accounts, store their secrets in a vault where the product supports it, and know in advance which accounts would need rotation. See the Secrets Management Guide and the Service Account Security Guide.

Detection

Detection has to work around the fact that the device itself may lie. Combine signals from the appliance, from the network around it and from identity systems behind it.

SignalWhere to lookWhat it can reveal
Vendor integrity check resultsIvanti external Integrity Checker Tool, Cisco core dump procedure in ED 25-03, Palo Alto Tech Support File analysisModified files and implants; CISA warns a clean ICT result is not sufficient on its own
Outbound connections from the applianceFirewall and NetFlow data for traffic sourced from the management or data plane to unknown hostsImplant beaconing, staging downloads, data exfiltration
Configuration driftScheduled config backups compared against a baselineNew local accounts, changed AAA or TACACS+ servers, new SSH keys, altered ACLs
VPN session anomaliesRemote syslog of VPN logs into the SIEMSame session ID from two IPs, logins for service accounts, logins outside normal geographies
Identity telemetry behind the VPNDomain controller events (4624, 4768, 4769), IdP sign-in logsBind and service accounts logging on interactively, Kerberos requests from the VPN pool at odd times
Log gapsSIEM ingestion health per deviceSilence from a device that normally sends logs, a common sign of tampering

Forward logs off-box in real time, because local logs are the first thing an attacker clears. When a vendor advisory says to collect forensic data (a Tech Support File for PAN-OS, a core dump for Cisco ASA), do it before rebooting or upgrading, since both destroy volatile evidence. See the Threat Hunting Guide.

Metrics that show the programme is working:

  • Time to mitigate KEV-listed edge CVEs: median hours from KEV addition or vendor advisory to patch or mitigation; ED 25-03 sets 48 hours for future updates as the federal benchmark.
  • Exposed management interfaces: count of appliance admin interfaces reachable from the internet (target zero).
  • Off-box log coverage: percentage of edge devices with real-time remote logging and an ingestion health alert.
  • Credential rotation readiness: percentage of appliances with a documented list of stored secrets and a tested rotation runbook.
  • End of support exposure: number of edge devices past vendor end of support.

Prevention and hardening

The hardening baseline lives in the Edge Device Security Guide. The decisions specific to exploitation risk are these.

  • Shrink the exposed surface: never expose management interfaces to the internet, and disable unused services such as SSL-VPN web portals that the business does not use. Fortinet noted that devices which never had SSL-VPN enabled escaped its April 2025 symlink persistence.
  • Treat KEV edge entries as emergencies: pre-agree a change window that allows patching or mitigating an internet-facing appliance within 48 hours, and an option to disconnect if no fix exists. See the Zero-Day Exploitation Playbook.
  • Retire end of support devices: CISA BOD 26-02 (February 2026) gives federal agencies 3 months to inventory, 12 months to decommission devices already out of support, and 18 months to replace all identified end of support edge devices. The same timeline is a reasonable benchmark for anyone. See the Edge Device End of Life Replacement Playbook.
  • Reduce what the device holds: use least-privilege bind accounts, short session lifetimes and phishing-resistant MFA on VPN logins, so stolen material is worth less.

Decision guidance: a small organisation with one firewall and no SOC gains most from a managed or cloud-delivered remote access service, strict auto-update and an MSP that can hit 48 hour mitigation. A large organisation with dozens of appliance families should focus on inventory accuracy, version telemetry and pre-approved emergency change, because the failure mode at scale is not knowing which devices run the affected build.

Response when a device is suspected

  1. Preserve evidence: collect the vendor-specified forensic bundle before reboot; if compromise is confirmed, isolate the device from the network without powering it off, as ED 25-03 instructs.
  2. Contain: fail over to a clean device or disconnect, and block the device's outbound traffic.
  3. Rebuild, do not just patch: export the configuration, factory reset (or the vendor's enhanced reset), install a fixed supported version, then review the configuration line by line before reimporting.
  4. Rotate and revoke: every stored password, API key, certificate and shared secret; domain accounts that authenticated through the device; Kerberos tickets; and active VPN and SSO sessions.
  5. Hunt behind the device: review identity logs for use of the exposed accounts since the earliest possible exploitation date, not since disclosure.
  6. Roll back safely: keep a known good image and configuration so a failed upgrade can be reverted without restoring the compromised build.

Framework mapping

FrameworkReferencesRelevance
NIST CSF 2.0ID.AM-01, ID.RA-01, PR.PS-02, PR.PS-03, DE.CM-01, RS.MI-01Inventory, vulnerability identification, maintenance and replacement, network monitoring, containment
NIST SP 800-53 Rev. 5CM-8, RA-5, SI-2, SA-22, SI-7, SC-7, AC-17, AU-6, IR-4Component inventory, scanning, flaw remediation, unsupported components, integrity, boundary and remote access, log review, incident handling
CIS Controls v8.1Control 1, Control 7, Control 12Asset inventory, continuous vulnerability management, network infrastructure management
ISO/IEC 27001:2022 Annex AA.8.8, A.8.20Management of technical vulnerabilities, network security
MITRE D3FENDD3-SU Software Update, D3-PH Platform Hardening, D3-SFV System Firmware Verification, D3-NTA Network Traffic Analysis, D3-ANCI Authentication Cache InvalidationCountermeasures matching the exploitation chain, including session and ticket revocation

Lessons from incidents

  • Patching without session revocation: CitrixBleed victims who patched but did not kill sessions saw attackers return with stolen cookies; the CitrixBleed 2 page shows the same lesson in 2025.
  • Forgetting the service accounts: Ivanti configuration dumps exposed the service account passwords used for authentication server configuration; organisations that reset user passwords only left the most privileged credential untouched.
  • Valid credentials on unpatched VPNs: the SonicWall SSL VPN account compromises of 2025 show that once credentials leak from an appliance, the VPN itself becomes the reuse path.

Practitioner checklist

  • Keep a live inventory of every internet-facing appliance with vendor, model, software version and end of support date.
  • Subscribe to the KEV feed and each edge vendor's PSIRT advisories, and route new edge entries straight to an on-call owner.
  • Agree in advance an emergency change process that can patch, mitigate or disconnect an edge device within 48 hours.
  • Remove internet exposure of all management interfaces and disable unused VPN and portal features.
  • Forward appliance logs off-box in real time and alert when a device stops sending them.
  • Document, per appliance, every stored credential, key, certificate and shared secret, with a tested rotation runbook.
  • Collect vendor forensic data before any reboot or upgrade of a suspected device.
  • On compromise, rebuild from a fixed image, rotate all secrets, revoke sessions and Kerberos tickets, and hunt from the earliest exploitation date.
  • Set a replacement deadline for end of support edge devices, using BOD 26-02 timelines as a benchmark.

Standards and references

Related NHI Mgmt Group resources: Edge Device Security Guide · Top 10 Ransomware Initial Access Vectors · Zero-Day and Exploited Vulnerabilities Guide · Vulnerability Management for Network Devices Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 27 October 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.