Security Advisory

CVE-2025-13086: OpenVPN HMAC Cookie Check Bypass

Last updated:
  • CVE-2025-13086

CVE-2025-13086 is a source-validation flaw in the OpenVPN server: an inverted memcmp() check accepts every HMAC cookie in the three-way handshake, so TLS sessions can be opened and server state consumed from IP addresses that never started a connection. OpenVPN 2.6.0 to 2.6.15 and 2.7_alpha1 to 2.7_rc1 are affected; NVD scores it 7.5.

Vulnerability summary

CVSS base score

7.5

High severity

EPSS score (FIRST)

0.7%

Probability of exploitation in the wild in the next 30 days

CISA KEV catalog

No

Not listed

Published

Vulnerability ID: CVE-2025-13086 · Source: NVD

Affected components

ComponentCPEVersion range
OpenVPN (community edition, server)cpe:2.3:a:openvpn:openvpn:*:*:*:*:community:*:*:*NVD: 2.6.0 to before 2.6.16, and 2.7_alpha1 to 2.7_rc1; fixed in 2.6.16 and 2.7_rc2, both released 17 November 2025 (OpenVPN announcement)
OpenWrt packages feed: openvpncpe:2.3:a:openvpn:openvpn:*:*:*:*:community:*:*:*2.6.14 until updated to 2.6.19 (master 2026-03-18, openwrt-24.10 2026-03-19); on 2026-11-18 openwrt-24.10 builds 2.6.23, openwrt-25.12 2.7.8, openwrt-23.05 2.5.8 (outside the range)
Buildroot openvpn packagecpe:2.3:a:openvpn:openvpn:*:*:*:*:community:*:*:*security bump to 2.6.16 on 2025-11-19

What does this mean for affected devices?

OpenVPN 2.6 added a stateless three-way handshake: before the server allocates a session, the client has to return an HMAC-based cookie, which proves that it really receives packets at its claimed source address. The OpenVPN announcement says that "due to a program code mistake all hmac cookies are accepted, thus breaking source IP address validation". As a consequence, "TLS sessions can be opened and state can be consumed in the server from IP addresses that did not initiate an initial connection." The fix also restricts accepted cookies to the current and two previous time slots, rejecting HMACs from a future slot. Joshua Rogers and SRLabs reported it; ZeroPath found it.

The 2.6.16 release notes put it plainly: the bug "renders the HMAC based protection against state exhaustion on receiving spoofed TLS handshake packets in the OpenVPN server inefficient". Affected are OpenVPN 2.6.0 through 2.6.15 and 2.7_alpha1 through 2.7_rc1. OpenVPN 2.6.16 and 2.7_rc2, both released on 17 November 2025, fix it. NVD's Primary score is CVSS v3.1 7.5 (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H), a network denial of service; OpenVPN as the CNA gives a CVSS 4.0 score of 4.6 with high attack complexity and names CWE-940, improper verification of source of a communication channel (NVD). NVD's description frames the impact as a denial of service for the client whose address is used.

Which devices are exposed to CVE-2025-13086?

Those that run OpenVPN 2.6 as a server: remote-access routers and machine-builder gateways that accept incoming tunnels on UDP 1194. The announcement and release notes describe the effect on the server; a device that only dials out as a client is not the target. Version 2.5 never had the code, and the Debian security tracker marks bullseye (2.5) as not affected because the vulnerable code is not present. Debian fixed bookworm (2.6.3-1+deb12u4) and trixie (2.6.14-1+deb13u1) in DSA-6069-1 and links the upstream fix commit for 2.6.16.

Which firmware build systems carry the fix?

Buildroot made a security bump to OpenVPN 2.6.16 on 2025-11-19, two days after the release. OpenWrt took four months: its packages feed built OpenVPN 2.6.14 with no patch for this CVE until the update to 2.6.19 on master on 2026-03-18 and on the openwrt-24.10 branch on 2026-03-19. Router images built from the feed in that window ship an affected OpenVPN. On 2026-11-18 the openvpn Makefile builds 2.6.23 on openwrt-24.10 and 2.7.8 on openwrt-25.12, both fixed, and 2.5.8 on openwrt-23.05, outside the affected range.

How do you fix CVE-2025-13086?

Upgrade the server to OpenVPN 2.6.16 or later on the 2.6 branch, or to 2.7.0 or later. OpenVPN names no configuration workaround, so a version or backport check is the test: read the version from the package database or the OpenVPN 2.6.x string in the binary, then look for the fix in the vendor's patch list.

Our OpenVPN and strongSwan firmware guide ranks this CVE with the other OpenVPN server flaws, including CVE-2026-35058, a tls-crypt-v2 crash fixed in 2.6.20, and the older CVE-2022-0547 deferred-auth plug-in bypass. On the IPsec side see CVE-2023-41913, CVE-2018-16151 and CVE-2026-78135 in strongSwan. The patch management guide covers rolling a fixed release out to a device fleet.

CVE-2025-13086 is not in the CISA KEV catalog as of 2026-11-18 (catalog version 2026.11.17). FIRST EPSS gives it 0.00665, the 52nd percentile, in the 2026-11-17 dataset (CISA KEV, FIRST EPSS).

How to fix this class of issueNo dedicated SoftWSP guide covers this weakness class; the fix is usually the vendor patch or firmware update referenced above. Our related guide explains how to plan and verify it: NIST SP 800-40 Rev. 4: enterprise patch management planning, explained.

Evidence

  • OpenVPN announcement: memcmp check for the HMAC verification in the 3-way handshake inverted; all HMAC cookies accepted, breaking source IP address validation; TLS sessions can be opened and state consumed from IP addresses that did not initiate a connection; fix also allows only [t-2;t] time slots; affected 2.6.0 to 2.6.15 and 2.7_alpha1 to 2.7_rc1; fixed in 2.6.16 and 2.7_rc2; reported by Joshua Rogers (found by ZeroPath) and SRLabs. — OpenVPN
  • OpenVPN release history: 2.6.16 and 2.7_rc2 released 17 November 2025; 2.6.16 security fix CVE-2025-13086 renders the HMAC based protection against state exhaustion on receiving spoofed TLS handshake packets in the OpenVPN server inefficient. — OpenVPN
  • NVD Primary CVSS 3.1 7.5 CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H; OpenVPN (CNA) CVSS 4.0 4.6; CWE-940 (CNA); CPE openvpn community 2.6.0 to <2.6.16 plus 2.7 alpha1 to rc1; published 2025-12-03. — NVD
  • Debian: bullseye not affected (vulnerable code not present); bookworm fixed 2.6.3-1+deb12u4 and trixie 2.6.14-1+deb13u1 (DSA-6069-1); unstable 2.7.0~rc2-2; upstream fix commit fa6a1824 (v2.6.16). — Debian security tracker
  • Buildroot commit 55d1f282 (2025-11-19) package/openvpn: security bump version to 2.6.16, fixes CVE-2025-13086. — Buildroot
  • OpenWrt packages: openvpn 2.6.14 -> 2.6.19 on master (e0ab7d0d, 2026-03-18) and openwrt-24.10 (a158e964, 2026-03-19), no Makefile change between 2025-10-02 and those commits; on 2026-11-18 openwrt-24.10 2.6.23, openwrt-25.12 2.7.8, openwrt-23.05 2.5.8. — OpenWrt packages feed
  • Not in the CISA KEV catalog, version 2026.11.17, checked 2026-11-18. — CISA KEV
  • EPSS 0.00665, percentile 0.52418, dataset 2026-11-17. — FIRST EPSS

Related advisories

Sources

  1. NVD
  2. OpenVPN security announcement: CVE-2025-13086, HMAC verification check
  3. OpenVPN release history (2.6.16 and 2.7_rc2)
  4. Debian security tracker: CVE-2025-13086
  5. Buildroot: openvpn security bump to 2.6.16
  6. OpenWrt packages: openvpn update to 2.6.19 (master)
  7. OpenWrt packages: openvpn update to 2.6.19 (openwrt-24.10)
  8. OpenWrt packages: openvpn Makefile
  9. CISA KEV
  10. FIRST EPSS

Frequently asked questions

Which OpenVPN versions are affected by CVE-2025-13086?

OpenVPN 2.6.0 to 2.6.15 and 2.7_alpha1 to 2.7_rc1. OpenVPN 2.6.16 and 2.7_rc2, released on 17 November 2025, fix it. OpenVPN 2.5 does not contain the vulnerable code (per the OpenVPN announcement, NVD and the Debian security tracker).

Does CVE-2025-13086 matter for a device that only runs OpenVPN as a client?

The OpenVPN announcement and release notes describe the effect on the server: it accepts handshakes, and keeps state, for source addresses that never started a connection. The devices to check first are routers and gateways that accept incoming OpenVPN tunnels.

Did OpenWrt router builds ship an affected OpenVPN?

Yes. The OpenWrt packages feed built OpenVPN 2.6.14, inside the affected range, until it moved to 2.6.19 on 2026-03-18 (master) and 2026-03-19 (openwrt-24.10). On 2026-11-18 those branches build fixed versions, and openwrt-23.05's 2.5.8 is outside the range.

Is CVE-2025-13086 in the CISA KEV catalog?

No, not as of 2026-11-18 (catalog version 2026.11.17). FIRST EPSS gives it 0.00665, the 52nd percentile, in the 2026-11-17 dataset (per CISA KEV and FIRST EPSS).