Executive brief
Sipwise rtpengine, a media proxy used in VoIP networks to handle audio and video traffic, contains a flaw in how it identifies where media packets are coming from. An attacker can exploit this to intercept private phone calls or inject their own audio into active conversations. This could lead to eavesdropping on sensitive communications or the disruption of telecommunications services.
Technical details
A vulnerability exists in the media-relay core of Sipwise rtpengine due to improper origin validation in its endpoint-learning logic. The software uses various learning modes (delayed, immediate, heuristic) to adapt to NAT environments, but in vulnerable versions, these modes fail to prevent unauthorized packets from triggering a source address update. Furthermore, when relaying SRTP, the engine fails to validate authentication tags, allowing unauthenticated packets to be processed. Attackers can exploit this to perform 'RTP Bleed' (redirecting media to an attacker-controlled host) or 'RTP Inject' (inserting arbitrary audio into a call) without being in a man-in-the-middle position. Version 13.4.1.1 mitigates this by limiting heuristic exposure to the first five packets and introducing a 'recrypt' flag to enforce SRTP authentication tag validation.
Affected products
- Sipwise rtpengine before 13.4.1.1
Timeline
- 2025-04-24: disclosed: First report to vendor
- 2025-07-03: patched: Vendor release 13.4.1.1
- 2025-08-01: advisory: Public disclosure and CVE assignment
References
- https://github.com/EnableSecurity/advisories/tree/master/ES2025-01-rtpengine-improper-behavior-bleed-inject
- https://github.com/sipwise/rtpengine/commits/rfuchs/security/
- https://github.com/sipwise/rtpengine/releases/tag/mr13.4.1.1
- https://www.openwall.com/lists/oss-security/2025/07/31/1
- http://seclists.org/fulldisclosure/2025/Aug/1
- http://www.openwall.com/lists/oss-security/2025/07/31/1
- https://lists.debian.org/debian-lts-announce/2026/07/msg00033.html