What a VPN protocol is and why the choice matters
A protocol is the agreement between your app and the server: how keys are exchanged, how packets are encrypted and wrapped, how a dropped connection is resumed. Your subscription usually includes several; the app picks one per connection, and switching is a menu option, not a commitment.
The choice shows up in three places. Throughput and battery drain differ because protocols cost different amounts of CPU per packet. Stability differs when you move between networks or sit behind strict firewalls. And visibility differs: some tunnels are trivial for network equipment to classify as VPN traffic, others are built to avoid classification. No protocol wins all three, which is why the menu exists.
WireGuard: speed and a small codebase
WireGuard is the default answer in 2026. Around four thousand lines of kernel code against the hundreds of thousands in older stacks, one modern cipher suite (ChaCha20-Poly1305) instead of a negotiation matrix, and a handshake that completes in a single round trip. In practice that means near-line-rate throughput on ordinary hardware, quick reconnects, and modest battery cost on phones.
It pays with two design traits. First, a standard WireGuard server assigns each peer a fixed address inside the tunnel, so the provider must add its own layer — rotating internal addresses, discarding peer state after disconnect — to avoid a stable per-user identifier; whether that is done is a provider property, not a protocol one. Second, WireGuard runs over UDP with a recognizable handshake pattern, so network equipment that wants to identify VPN traffic finds it easy to spot, and networks that drop unknown UDP break it entirely.
OpenVPN: maturity and flexibility
OpenVPN has been in production since the early 2000s, and that age is its asset: two decades of scrutiny, support on virtually every platform and router, and configuration options for almost any network layout. It runs over UDP for speed or TCP for reachability, and on TCP port 443 it at least uses the port firewalls rarely close — though its TLS handshake still differs from a browser's, so classification remains possible.
The cost is weight. OpenVPN processes packets in user space, crossing the kernel boundary on every packet, and its throughput on the same hardware is typically a fraction of WireGuard's, with reconnects measured in seconds rather than instantly. The codebase and its OpenSSL dependency are orders of magnitude larger, which means a bigger audit surface. It remains the right tool on legacy routers, in corporate setups that require certificates, and on networks where UDP is unusable.
IKEv2/IPsec: built for phones
IKEv2's distinguishing feature is MOBIKE, a standard extension that lets an established tunnel survive a change of network address. Walk out of Wi-Fi coverage onto LTE and the session migrates without a new handshake — the exact event that interrupts other protocols mid-call. Native support in iOS and Android, with encryption handled in the operating system's IPsec stack, keeps battery cost low and makes it a common default in mobile apps.
Its weaknesses are environmental. IKEv2 needs UDP ports 500 and 4500, which restrictive networks — hotel Wi-Fi, office firewalls — often block, and it has no TCP fallback of its own. Configuration is also less transparent than WireGuard's: IPsec parameters are negotiated, and a poorly configured server can settle on weaker settings than intended. Strong choice on mobile, brittle on hostile networks.
Xray/VLESS: a tunnel that looks like HTTPS
VLESS is a lightweight transport protocol from the Xray project: minimal framing, no built-in encryption of its own, designed to run inside a real TLS session. Because the encryption layer is standard TLS — the same protocol every website uses — a VLESS connection on port 443 presents an ordinary certificate handshake followed by an opaque encrypted stream, which is exactly what any HTTPS session looks like from outside.
This matters on networks that classify and throttle or drop recognizable VPN protocols. Where WireGuard's handshake or OpenVPN's TLS quirks give the tunnel away, a well-configured Xray endpoint gives a classifier no distinguishing pattern beyond statistics. The costs: more moving parts on the server, throughput bounded by TLS processing rather than kernel networking, and a small ecosystem of clients compared with mainstream protocols. Aurora runs WireGuard for everyday speed and Xray/VLESS on its own nodes for networks where standard tunnels are unreliable.
Which protocol should you choose
Start with WireGuard and change only when something specific breaks. The practical mapping:
- Default, speed, gaming, large downloads — WireGuard
- Phone that hops between Wi-Fi and mobile data all day — IKEv2, or WireGuard if reconnects feel fast enough
- Strict firewall that allows only web traffic — OpenVPN over TCP 443, or better, Xray/VLESS
- Network where recognizable VPN protocols stall or drop — Xray/VLESS over TLS
- Old router or a corporate certificate requirement — OpenVPN
- Not sure — leave the app on automatic and let it fall back in this order