Engineer working on a laptop connected to a device test rig, surrounded by wiring harnesses and test hardware

Firmware vulnerabilities & fixes

pppd vulnerabilities in embedded firmware: find the version, rank the EAP bugs, patch or backport

Last updated:

The pppd vulnerabilities in embedded firmware that matter most as of 2026-11-16 are CVE-2020-8597, an EAP stack overflow in ppp 2.4.2 through 2.4.8 that an unauthenticated link peer can trigger even when EAP was never negotiated, and the 2026 fixes in ppp 2.5.4: a PEAP response overflow (CVE-2026-75883) and a pre-authentication LCP overflow. Find the pppd version and its compiled-in authentication methods, then upgrade or backport the fixes and drop plugins and methods the product does not use.

Key takeaways

  • Industrial cellular routers are a typical example.
  • Five CVEs and one 2026 advisory cover most of what a firmware scan of a pppd image turns up, and their ranges are worth reading against the upstream code, because NVD's are incomplete for two of them.
  • CVE-2020-8597 is a bounds check that could never be true.
  • ppp 2.5.4, released on 2026-09-15, is mainly a security release.

Where does pppd ship in embedded devices?

pppd ships wherever a device brings up a link with the Point-to-Point Protocol: DSL routers and CPE that dial PPPoE or PPPoA to the ISP, cellular and LTE routers and modems that still run PPP over a serial or USB AT-command port, industrial gateways with dial-up or cellular backup, and VPN appliances that terminate PPTP or L2TP. The ppp project describes itself as Paul's PPP Package for Linux and Solaris, and on 2026-11-16 lists 2.5.5, released on 31 October 2026, as the latest version.

Industrial cellular routers are a typical example. Siemens' SSA-809841 fixed CVE-2020-8597 in SCALANCE M-800 and S615 and in the RUGGEDCOM RM1224 LTE router, all versions before V6.3, and notes that the PPP function is off by default but used for dial-in and point-to-point connections, where "the vulnerability could be exploited by a malicious peer". NETGEAR's PSV-2020-0136 fixed the same flaw in its WAC510 access point in firmware 8.2.6.7.

Most images get pppd from a build system package:

  • OpenWrt. The ppp package Makefile on the main branch builds 2.5.5 with --disable-peap and --disable-eaptls, and splits PPPoE, PPPoA, PPTP, L2TP, RADIUS and pppdump into separate packages. The release branches differ: openwrt-25.12 carries 2.5.2, openwrt-24.10 carries 2.5.1 (also with PEAP off), and openwrt-23.05 carries a 2.4.9 snapshot dated 2021-01-04.
  • Buildroot. The pppd package builds 2.5.5 on master and passes --enable-eaptls --enable-peap whenever OpenSSL is in the image; its Config.in selects OpenSSL automatically on glibc toolchains.
  • Yocto. OpenEmbedded-Core's ppp recipes carry 2.5.5 on master, 2.5.2 on walnascar and whinlatter, and 2.5.0 on the scarthgap LTS branch, built with OpenSSL by default.

Our embedded Linux build systems and SBOM guide explains how those recipes surface in an SBOM. pppd usually sits next to a BusyBox userland and links against the C library, so our glibc, uClibc and musl guide covers the layer underneath. Small RTOS devices may instead use the PPP stack in lwIP, which is derived from pppd code.

Which pppd CVEs matter most for device makers?

Five CVEs and one 2026 advisory cover most of what a firmware scan of a pppd image turns up, and their ranges are worth reading against the upstream code, because NVD's are incomplete for two of them. CVSS below is NVD's own score unless marked; EPSS values are FIRST's for 2026-11-15, and none of these CVEs is in CISA's KEV catalog version 2026.11.13.

CVE or advisory What it is Affected Fixed upstream CVSS EPSS
CVE-2020-8597 Stack overflow of the peer-name buffer in EAP request and response handling, reachable before authentication 2.4.2 through 2.4.8 Commit 8d7970b8 (2020-02-03), release 2.4.9 9.8 0.199
CVE-2015-3310 Overflow in the RADIUS plugin's session-ID builder when the pppd PID exceeds 65535, crashing pppd 2.4.6 and earlier per NVD; upstream code unchanged until 2.4.9 Commit 858976b1 (2020-01-04), release 2.4.9 4.3 (v2) 0.054
CVE-2022-4603 Array index check in the offline pppdump tool Before 2.5.0 Commit a75fb7b1, release 2.5.0 6.5 0.009
CVE-2024-58250 passprompt plugin mishandles privileges Before 2.5.2 Plugin removed in 2.5.2 (2024-12-31) 9.3 (CNA) 0.002
CVE-2026-75883 PEAP response copies a TLS record of up to 16384 bytes into a fixed buffer 2.5.0 through 2.5.3 2.5.4 (2026-09-15) 6.8 (CNA) 0.003
GHSA-frhv-j822-2pvx Pre-authentication overflow of the LCP Configure-NAK buffer 2.5.3 and earlier 2.5.4 (2026-09-15) 4.2 (GitHub) n/a

The project's published advisories list nine GitHub advisories from August and September 2026, most without a CVE. Several, including RADIUS plugin overflows that date back to 2.4.2, only matter where pppd is installed setuid-root and an unprivileged user can start it, which is rare on a single-purpose device but worth checking. The LCP advisory names a CVE ID, but on 2026-11-16 CVE.org had no published record for it, so it is listed here by its GitHub identifier.

How does the EAP overflow (CVE-2020-8597) work?

CVE-2020-8597 is a bounds check that could never be true. CERT/CC's note VU#782301, published on 2020-03-04, explains that when an EAP packet carries the MD5-Challenge type, pppd checks an embedded 1-byte length against the packet length and then copies the peer's hostname into a stack buffer, but the check is wrong and allows a copy of arbitrary length. The upstream fix, committed by Paul Mackerras on 2020-02-03 and credited to Ilja Van Sprundel of IOActive, changes one comparison in each of eap_request() and eap_response() from vallen >= len + sizeof (rhostname) to len - vallen >= sizeof (rhostname).

Turning EAP off does not close it. CERT/CC describes a second logic flaw: eap_input() did not check whether EAP had been negotiated during the Link Control Protocol phase, so a peer could send an EAP packet even after pppd refused EAP, and "it is incorrect to assume that pppd is not vulnerable if EAP is not enabled". CERT/CC lists no viable workaround, and because pppd usually runs as root, a successful exploit can run code as root. The note also covers lwIP: its default packages are not affected, but builds that enable EAP at compile time are.

The attacker is whoever sits on the other end of the PPP link, which on a device means several paths:

  • PPPoE on the WAN port. Anyone who can answer PPPoE discovery on the WAN segment, such as a rogue access concentrator on a shared or compromised access network, becomes the peer.
  • Cellular and serial links. A modem that runs PPP to a carrier or a fake base station, or a serial dial-in port on an industrial gateway, gives the peer role to whatever answers the call.
  • VPN servers. A device that terminates PPTP or L2TP runs pppd as a server for any client that connects, which is why CERT/CC's proof of concept targets PPTP servers.

As of 2026-11-15, FIRST's EPSS gives CVE-2020-8597 0.199, the 97th percentile of all scored CVEs, although it is not in CISA's KEV catalog. Our KEV and EPSS prioritisation guide explains how to weigh those two signals.

What changed in ppp 2.5.4 and 2.5.5?

ppp 2.5.4, released on 2026-09-15, is mainly a security release. Its release notes say the fixes range "in severity up to 6.8 (moderate)" and that many matter only when pppd is setuid-root. Two are reachable from the link peer.

The PEAP flaw, GHSA-rwr9-4vx8-vc35 (CVE-2026-75883), exists from 2.5.0, when PEAP support arrived, through 2.5.3. A server that requests PEAP can make pppd copy a TLS record of up to 16384 bytes into a fixed global buffer, corrupting data after it or crashing pppd. pppd is only exposed if it agrees to PEAP, which requires the ca or capath option, and the advisory gives ./configure --disable-peap as a build-time workaround. NVD lists the record as awaiting analysis, with no affected-version configuration, on 2026-11-16.

The LCP flaw, GHSA-frhv-j822-2pvx, is a pre-authentication overflow of the 1500-byte buffer pppd uses to build an LCP Configure-NAK. Each rejected 4-byte authentication-type option is answered with a 5-byte suggestion and duplicates are not rejected, so one legal Configure-Request with 301 or more repeated PAP options overflows it. The precondition is a CHAP-only configuration with refuse-pap and refuse-eap, which the advisory calls "a legitimate, commonly recommended hardening"; the default configuration does not reach this path.

2.5.4 also makes all RADIUS plugin options and almost all EAP-TLS and PEAP options privileged, removes EAP-SRP, refuses world-readable secrets files by default and lets pppd run as non-root with CAP_NET_ADMIN. 2.5.5, released on 2026-10-31, is a bug-fix and security release that fixes validation of received frames in the RADIUS plugin and closes cases where an MPPE link could come up without the encryption the configuration required.

Do the older pppd CVEs still matter?

The older pppd CVEs matter less than the EAP flaw, but each has a detail that trips up scanners. CVE-2015-3310 is a crash in the RADIUS plugin's rc_mksid(), which formats the time, PID and a counter into a 15-byte buffer that overflows when the PID is greater than 65535. NVD lists ppp 2.4.6 and earlier as affected, and the Debian tracker shows Debian's 2015 fix in 2.4.6-3.1. Upstream, though, the 2.4.8 source still carries the 15-byte buffer: the fix commit landed on 2020-01-04 and first shipped in 2.4.9, so an image with an unpatched upstream 2.4.7 or 2.4.8 build and the RADIUS plugin is affected even though NVD's range says otherwise. It only matters on devices that load radius.so, usually access concentrators or VPN servers.

CVE-2022-4603 is in pppdump, an offline tool that decodes recorded PPP sessions. NVD's own description notes that pppdump is not used to set up a connection, is not installed setuid-root and is never run automatically, and the record says the real existence of the issue is doubted; Debian rates it unimportant. On OpenWrt pppdump is a separate package, so it is often absent. The fix shipped in 2.5.0.

CVE-2024-58250 covers the passprompt plugin, which the project removed in 2.5.2 rather than repairing. The CNA's 9.3 score assumes local access, so on a device it matters only if the plugin is installed and something can start pppd with it. The Debian tracker marks bookworm and bullseye as ignored because removing the plugin could break configurations, while Yocto's scarthgap recipe carries the removal as a backport patch. Older pppd releases also have CVE-2014-3158, an options-file parsing overflow fixed in 2.4.7.

How do you find the pppd version and build options in a firmware image?

Start with the version, then establish which authentication methods and plugins the build contains, because the 2026 fixes depend on configuration as much as on version.

  1. Plugin directorypppd loads plugins from a directory named after its version, so /usr/lib/pppd/2.4.9/ or /usr/lib/pppd/2.5.1/ in the unpacked root file system gives the version directly; OpenWrt installs every plugin package there.
  2. Binary stringsSearch the pppd binary for the version string. Each plugin also exports a pppd_version symbol, and pppd refuses a plugin built for another version, so a mismatch points to a partially updated image.
  3. Package metadataOpenWrt's opkg database records ppp or ppp-multilink with versions such as 2.5.1-1, or 2.4.9.git-2021-01-04-4 on 23.05, and Buildroot and Yocto record the version in their manifests.
  4. Startup logOn a device you control, pppd logs pppd <version> started by at startup.
  5. EAP-TLS and PEAPSearch the binary for EAP-TLS and PEAP strings, or check whether pppd links against libssl or libcrypto. A pppd without OpenSSL cannot have the PEAP code, because 2.5.x requires --disable-eaptls and --disable-peap when OpenSSL is missing.
  6. Plugins and optionsList radius.so, pppoe.so, pptp.so and similar plugins, and read /etc/ppp/options, files under /etc/ppp/peers and any scripts that build pppd command lines at boot. Look for refuse-pap, refuse-eap, require-chap, ca, capath and plugin lines.
  7. Setuid bitCheck whether pppd is installed setuid-root; several 2.5.4 advisories depend on it.
  8. Vendor patchesCheck the vendor patch list or the distribution changelog for backported fixes before calling a CVE open.

How do you mitigate pppd exposure on a device?

Patching is the only real fix for the EAP overflow, but the build and configuration decide how much else is reachable. As the pppd(8) manual describes them, refuse-eap and refuse-pap stop pppd authenticating itself with those methods and require-chap or require-eap set what it demands of the peer. Use them to narrow what the product accepts, but keep three limits in mind:

  • Refusing EAP does not stop CVE-2020-8597 on unpatched 2.4.x builds, as CERT/CC states.
  • CHAP-only hardening reaches the LCP flaw on 2.5.3 and earlier, so that configuration needs 2.5.4 or a backport.
  • Build-time removal beats configuration. Configure with --disable-peap and --disable-eaptls if the product never uses them, as OpenWrt does, and drop the RADIUS, PPTP and pppdump packages unless the product needs them. Code that is not compiled in cannot be reached by a later configuration change.

Also keep pppd off setuid-root on devices where only a root service starts it, keep secrets files readable by root only, and never let a WAN-side service such as a web interface pass user input into pppd options. Our hard-coded credentials guide covers per-device PPPoE and dial-in secrets.

How does a device maker patch or backport pppd?

  1. Upgrade where you canAs of 2026-11-16 the current release is 2.5.5. The 2.5.0 release notes describe a move to GNU Autoconf and Automake and breaking changes to the plugin API, and later 2.5 releases make more options privileged, so rebuild vendor plugins and test dial-up, PPPoE and VPN scripts after the move.
  2. Backport the EAP fix on 2.4.xApply commit 8d7970b8 and the companion commit 8d45443b, which makes pppd ignore EAP messages when it has not requested or agreed to EAP. The first is a two-line change that is small enough to carry on old trees: the Debian tracker for CVE-2020-8597 shows it fixed in 2.4.6 for jessie (DLA-2097-1) and 2.4.7 for stretch and buster (DSA-4632-1).
  3. Backport the rest of 2.4.9If you stay on 2.4.7 or 2.4.8, also apply the rc_mksid() fix for CVE-2015-3310 when the RADIUS plugin ships.
  4. Backport the 2026 fixes on 2.5.0 to 2.5.3Take the PEAP and LCP fixes from 2.5.4, or build with PEAP disabled. On 2026-11-16 Debian's tracker for CVE-2026-75883 marks bookworm's 2.4.9 not affected and trixie's 2.5.2 vulnerable, and Yocto's scarthgap recipe for 2.5.0 builds with OpenSSL and carries no patch for it, so a scarthgap-based image needs its own backport or a --disable-peap build.
  5. Ship and documentDeliver the update through a signed update path, as our firmware OTA update security guide describes, regenerate the SBOM, and state per CVE whether the product is affected. Our CSAF and VEX for device makers guide shows the format, for example "not affected" for CVE-2026-75883 on a build without PEAP.

Worked example: how would you triage pppd in an LTE router image?

Synthetic example, not a real product. A router maker scans an image for an LTE router with a DSL fallback, built with Buildroot on a glibc toolchain.

  1. Confirm the baseThe plugin directory is /usr/lib/pppd/2.5.2/, the pppd binary carries the same version string, and the vendor patch list shows no ppp backports.
  2. Check the buildpppd links against libssl, matching Buildroot's behaviour of enabling EAP-TLS and PEAP when OpenSSL is present, and pppoe.so is the only plugin installed.
  3. Match the CVEsCVE-2020-8597, CVE-2015-3310, CVE-2022-4603 and CVE-2024-58250 do not apply to 2.5.2. CVE-2026-75883 is in range and PEAP is compiled in, but no options file sets ca or capath, so pppd refuses PEAP; it is recorded as present in code but not reachable in this configuration.
  4. Check the LCP flawThe PPPoE peer file sets refuse-pap and refuse-eap to force CHAP, which meets GHSA-frhv-j822-2pvx's precondition, and the peer is the ISP's access concentrator or anyone who can impersonate it on the DSL segment.
  5. FixShip an update to 2.5.5 built with --disable-peap and --disable-eaptls, then rebuild, re-scan and publish a VEX statement per CVE and advisory.

How does AI-assisted analysis help with pppd findings?

SoftWSP's automated scan unpacks a firmware image, catalogs the software packages in the extracted file system with their versions (the inventory our firmware SBOM guide describes), and matches each one against CVE records, CISA's KEV catalog and FIRST's EPSS scores. This is deterministic code: the code scores, and AI narrates. You can inspect the resulting CycloneDX file in the free SBOM viewer.

Version matching follows NVD's records and inherits their limits, which are significant for pppd. NVD files most pppd CVEs under point-to-point_protocol_project:point-to-point_protocol, the identity OpenWrt and Buildroot declare, but CVE-2022-4603 under samba:ppp, and on 2026-11-16 CVE-2026-75883 had no affected-version configuration at all. NVD's range for CVE-2015-3310 stops at 2.4.6, below the upstream fix. A package version also cannot show whether PEAP is compiled in, whether refuse-pap and refuse-eap are set or whether a vendor backported the EAP fix. AI's part is the written explanation of what matched, citing the component, version and record behind each sentence. An engineer reviews every finding before the report is delivered, and that review is where the second CPE identity, the project's GitHub advisories, the compiled-in methods, the options files and any backport become a documented "affected" or "not affected" decision.

pppd often runs next to other network daemons on the same device; see our hostapd and wpa_supplicant guide for the Wi-Fi authentication side and our OpenVPN and strongSwan guide for VPN daemons, or device & firmware security services to have a specific image reviewed.

This guide draws on the ppp project's home page, GitHub releases, published security advisories and fix commits, the pppd(8) manual for 2.5.5, CERT/CC VU#782301, Siemens SSA-809841, NETGEAR PSV-2020-0136, the Debian Security Tracker, the OpenWrt, Buildroot and OpenEmbedded-Core ppp packages, NVD, CVE.org, CISA's KEV catalog (version 2026.11.13) and FIRST EPSS data dated 2026-11-15, all read directly on 2026-11-16.

Put this guide to work

Frequently asked questions

Which pppd vulnerabilities matter most for embedded devices?

As of 2026-11-16, start with CVE-2020-8597, a stack buffer overflow in pppd's EAP code that affects ppp 2.4.2 through 2.4.8. NVD scores it 9.8, CERT/CC note VU#782301 says an unauthenticated peer can trigger it even when EAP was never negotiated, and FIRST's EPSS puts it in the 97th percentile. It is not in CISA's KEV catalog. Next come the 2026 fixes in ppp 2.5.4: CVE-2026-75883, a PEAP response overflow in 2.5.0 to 2.5.3, and a pre-authentication LCP overflow published as GHSA-frhv-j822-2pvx.

How do I find the pppd version in a firmware image?

Look at the plugin directory first: pppd installs its plugins under a directory named after its version, such as /usr/lib/pppd/2.4.9, and every plugin carries a pppd_version string that must match the daemon. The pppd binary also contains the version string, and pppd logs 'pppd <version> started by' at startup. Package records give the same number: OpenWrt's opkg database lists ppp, and Buildroot and Yocto record it in their build manifests.

Does refusing EAP protect pppd against CVE-2020-8597?

No. CERT/CC's note VU#782301 states that it is incorrect to assume pppd is not vulnerable if EAP is not enabled or not negotiated, because the vulnerable eap_input code did not check whether EAP had been agreed during LCP. CERT/CC lists no viable workaround. The fix is the bounds-check patch upstream (commit 8d7970b8), ideally with the companion commit that makes pppd ignore EAP packets when it is not doing EAP, or ppp 2.4.9 or later.

Is a device on ppp 2.5.x safe?

Not automatically. ppp 2.5.0 added PEAP support, and CVE-2026-75883 affects 2.5.0 through 2.5.3 when pppd is configured to accept PEAP. ppp 2.5.4, released on 2026-09-15, fixes it together with an LCP overflow and several RADIUS plugin issues, and 2.5.5 followed on 2026-10-31 with further RADIUS frame validation. Whether a given build is exposed depends on its configure options: OpenWrt builds with PEAP disabled, while Buildroot enables PEAP whenever OpenSSL is in the image.

Does SoftWSP find pppd in my firmware automatically?

SoftWSP's automated scan catalogs the software packages in an unpacked firmware file system with their versions and matches them against CVE records, CISA's KEV catalog and FIRST EPSS scores. AI then writes a plain-language summary that cites the component, version and record behind each finding, and an engineer reviews it. For pppd that review includes mapping the package to both of NVD's product identities for ppp, checking the project's GitHub advisories for issues NVD has not mapped to a version range, and checking which authentication methods and plugins the build actually contains.

Sources

  1. ppp project home page (latest release)
  2. ppp-project/ppp: GitHub releases
  3. ppp-project/ppp: published security advisories
  4. ppp-project/ppp commit 8d7970b8: pppd: Fix bounds check in EAP code (CVE-2020-8597)
  5. ppp-project/ppp commit 8d45443b: pppd: Ignore received EAP messages when not doing EAP
  6. ppp-project/ppp commit 858976b1: radius: Prevent buffer overflow in rc_mksid() (CVE-2015-3310)
  7. ppp-project/ppp commit a75fb7b1: pppdump bounds fix (CVE-2022-4603)
  8. ppp-project/ppp commit 0a66ad22: pppd: Remove passprompt plugin (CVE-2024-58250)
  9. GHSA-rwr9-4vx8-vc35: PPPD buffer overflow in PEAP response code (CVE-2026-75883)
  10. GHSA-frhv-j822-2pvx: pre-authentication LCP Configure-NAK buffer overflow
  11. GHSA-626q-hc9r-2crc: PPPD radius plugin user-triggerable buffer overflows
  12. CERT/CC Vulnerability Note VU#782301: pppd vulnerable to buffer overflow due to a flaw in EAP packet processing
  13. Siemens SSA-809841: Buffer Overflow Vulnerability in Third-Party Component pppd
  14. NETGEAR PSV-2020-0136: Unauthenticated Remote Buffer Overflow Attack in PPPD on WAC510
  15. Debian Security Tracker: CVE-2020-8597
  16. Debian Security Tracker: CVE-2015-3310
  17. Debian Security Tracker: CVE-2024-58250
  18. Debian Security Tracker: CVE-2026-75883
  19. OpenWrt: ppp package Makefile
  20. Buildroot: pppd package makefile
  21. OpenEmbedded-Core: ppp recipes (master and scarthgap)
  22. ppp 2.5.5 manual page: pppd(8)
  23. NIST NVD: CVE-2020-8597
  24. NIST NVD: CVE-2015-3310
  25. NIST NVD: CVE-2022-4603
  26. NIST NVD: CVE-2024-58250
  27. NIST NVD: CVE-2026-75883
  28. CISA: Known Exploited Vulnerabilities (KEV) Catalog
Was this guide helpful?

In this cluster

pppd vulnerabilities in embedded firmware: find the version, rank the EAP bugs, patch or backport