Voice over Wi-Fi (VoWiFi), better known to subscribers as Wi-Fi calling, lets a phone place and receive calls over any Wi-Fi network as if it were attached to the mobile network. It is convenient for the subscriber, it extends coverage indoors at no cost to the operator, and it is switched on by default on most modern handsets. What is far less visible is that VoWiFi opens a path from the public internet straight into the operator voice core. That path terminates on a single network function, the ePDG, and the way that function is built and configured decides how safe the whole arrangement is.
This article explains how VoWiFi actually works, where its security breaks down, what recent public research found in deployed networks, and how operators can harden it. For the wider voice picture, see our explainer on VoLTE and our review of IMS and SIP security.
What VoWiFi is and how it works
VoWiFi carries the same IMS voice service as VoLTE, but the access leg is untrusted Wi-Fi and the open internet rather than the licensed radio network. The handset builds a secure tunnel from itself, across whatever Wi-Fi and internet path it happens to be on, into the operator core. 3GPP specifies this as untrusted non-3GPP access to the Evolved Packet Core in TS 23.402.
The tunnel terminates on the Evolved Packet Data Gateway (ePDG). The interface between the handset and the ePDG is called SWu, and the tunnel itself is an IPsec tunnel established with the Internet Key Exchange version 2 protocol (IKEv2, RFC 7296). Setup happens in two stages. The first stage negotiates the cryptographic algorithms and performs a Diffie-Hellman key exchange to establish keys. The second stage, protected by those keys, authenticates the subscriber using EAP-AKA with the credentials on the SIM. Once the tunnel is up, SIP signalling and RTP media flow through it to the IMS core exactly as they would over VoLTE.
Why the ePDG is a security-sensitive box
The ePDG is unusual among core network functions because it is internet-facing by design. It has to be reachable from any Wi-Fi network anywhere in the world, which means it is reachable by anyone. It terminates the IPsec tunnel for every Wi-Fi-calling subscriber, and it bridges a fundamentally untrusted access network into the trusted voice core. Anything that degrades the ePDG degrades voice availability, and anything that weakens its tunnel setup weakens the confidentiality of calls. In short, it carries the exposure of an internet gateway while sitting on the edge of the crown jewels.
Where VoWiFi security breaks down
Weak Diffie-Hellman groups and silent downgrade
IKEv2 negotiates its key exchange from a set of predefined Diffie-Hellman groups, and several of the older groups are weak. The 768-bit and 1024-bit MODP groups have been considered breakable by well resourced actors for years. Public research presented at USENIX Security 2024, which scanned commercial VoWiFi deployments worldwide, found that a number of production ePDGs still offered these weak groups, and that a widely deployed baseband could fall back to weak, unannounced modes. When the key exchange is weak, the confidentiality of the entire IPsec tunnel is undermined regardless of how strong the later stages are. The fix is to offer only strong groups, such as 2048-bit MODP or 256-bit elliptic curve and above, and to disable any weak fallback on both the network and the device profile.
Subscriber identity exposure during setup
Authentication uses EAP-AKA, which carries a subscriber identity during the second stage of the exchange. If the deployment permits the permanent identity to appear rather than a privacy pseudonym, an on-path attacker on the untrusted Wi-Fi segment can capture it. This is the VoWiFi equivalent of the classic identity-exposure problem, and it matters because the permanent identity is the key to tracking a subscriber. For background on why that identifier is so sensitive, see what IMSI is and why it matters, and for the wider ciphering picture see our review of encryption across mobile generations.
Source address and location leakage
Because the tunnel endpoint is the subscriber's real public IP address, and because ePDG selection is frequently driven by DNS on a per-country basis, VoWiFi can leak coarse location and presence information about where a subscriber is and which home network they belong to. That is a privacy exposure even when the call content itself is protected, and it connects to the broader family of location tracking attacks against mobile networks.
An internet-facing attack surface
An exposed ePDG can be discovered, fingerprinted, and targeted like any other internet host. IKE initiation floods and malformed-packet attacks can exhaust its capacity or crash implementations that do not validate input robustly, producing a localised denial of voice service. The ePDG needs the same perimeter discipline as any internet gateway: rate limiting, filtering, and strict input validation.
Weak ciphers and misconfiguration
Beyond the key exchange, some deployments negotiate weak or outdated ESP encryption and integrity algorithms, and a small number have historically allowed null encryption in edge cases. The tunnel is only as strong as the weakest algorithm both ends will accept, so the negotiated cipher suite has to be constrained deliberately rather than left to defaults.
What operators can do
Offer only strong Diffie-Hellman groups on the ePDG and disable weak fallback, on both the network and the handset profile where the operator controls it. Use privacy identities during EAP-AKA so the permanent subscriber identity never appears on the untrusted segment. Constrain the IPsec cipher suite to strong ESP algorithms and never permit null encryption. Treat the ePDG as an internet perimeter: harden it, rate-limit IKE, and validate input. Monitor IKE and IPsec setup for anomalies with telecom-aware detection rather than assuming the tunnel is a black box, an approach that pairs naturally with telecom intrusion detection. Finally, treat VoWiFi as a first-class part of the IMS attack surface, not a convenience feature bolted on the side, and include it in IMS and SIP security assessments.
Key takeaways
- VoWiFi carries IMS voice into the core from untrusted Wi-Fi through the ePDG over an IKEv2/IPsec tunnel, so the ePDG is an internet-facing function on the edge of the voice core.
- Weak Diffie-Hellman groups and silent downgrade are the most impactful issue, and public research in 2024 confirmed weak groups are still offered in production.
- Identity exposure during EAP-AKA and source-IP location leakage are real privacy risks even when call content is protected.
- The ePDG needs internet-perimeter hardening against scanning and denial of service.
- The practical defence is strong crypto configuration plus telecom-aware monitoring, treating VoWiFi as part of the IMS attack surface.
If you want a review of your VoWiFi and IMS exposure, or of the wider voice core around it, get in touch at [email protected].
