Which VPN Protocol Should You Use?
Five questions, then a straight answer — including “you do not need the stealthy one”.
Use
WireGuard
The modern default. A small, audited codebase using fixed modern cryptography, fast enough that it rarely becomes the bottleneck.
- Nothing on your network is blocking VPN traffic, so there is no reason to pay for obfuscation. WireGuard is the modern default and will usually not be your bottleneck.
Second choice · OpenVPN
Keep OpenVPN over TCP/443 in reserve for hotel and conference networks that permit web traffic and nothing else. It is slower, and on those networks it is often the only thing that connects.
On JaguarVPN
Aurora is the JaguarVPN default and the right answer here — sized for full line speed, with fresh keys per session and a kill switch that fails closed. MIRAGE, the stealth option, would cost you speed for a problem you do not have.
How Aurora works →Compare all of them side by side
WireGuard
The modern default. A small, audited codebase using fixed modern cryptography, fast enough that it rarely becomes the bottleneck.
- Consistently the fastest mainstream option, and quick to reconnect after a dropped link
- Around 4,000 lines of code, which is small enough to have been reviewed properly
- Excellent battery behaviour on phones — it is idle when you are
- Trivially recognisable to deep-packet inspection: it has a distinct handshake and no cover story
- UDP only, so a network that permits nothing but TCP/443 blocks it outright
- Assigns each client a stable internal address, which providers must design around rather than ignore
Skip it when the network you are on actively looks for and blocks VPN traffic.
OpenVPN
The twenty-year incumbent. Slower and heavier than WireGuard, but it runs over TCP, which is occasionally the only thing that matters.
- Runs on effectively every platform, router firmware and appliance in existence
- Over TCP port 443 it can cross networks that permit only web traffic
- Extensively audited, and understood by every network administrator
- Noticeably slower than WireGuard, particularly on modest hardware
- TCP mode suffers meltdown: a reliable protocol inside another reliable protocol retransmits twice over, and throughput collapses on a lossy link
- Still identifiable to inspection — running on port 443 is not the same as looking like HTTPS
Skip it when you have a WireGuard-capable client and nothing is blocking UDP.
IKEv2 / IPsec
The mobile specialist. Its defining feature is surviving a change of network without dropping the tunnel.
- MOBIKE keeps the session alive as you move between Wi-Fi and mobile data
- Built into iOS, macOS and Windows, so it needs no third-party client
- Fast, and reconnects almost instantly after a tunnel interruption
- Uses fixed, well-known UDP ports (500 and 4500) that restrictive networks block first
- No obfuscation whatsoever
- Implementation quality varies widely between vendors
Skip it when you need to get through a network that filters by port.
An obfuscated protocol
A category rather than one protocol: designs whose goal is that traffic analysis finds nothing worth blocking. This is what you need where VPNs themselves are the thing being filtered.
- Passes networks that detect and drop conventional VPN traffic
- The better designs also survive active probing, where a censor connects to a suspected server to see how it answers
- Usually the only thing that works under national-level filtering
- Slower, sometimes dramatically — camouflage costs round trips and overhead
- More moving parts, so more that can go wrong
- Unnecessary on any network that is not filtering, where it is pure cost
Skip it when nothing on your network is blocking VPNs — the speed cost buys you nothing.
PPTP and L2TP/IPsec
Obsolete. Listed so that the answer to "should I use PPTP?" is unambiguous.
- Present in very old equipment
- PPTP’s encryption has been practically breakable since 2012 and offers no meaningful protection
- L2TP/IPsec is slow, doubly encapsulated, and blocked almost everywhere
- Both are widely disabled or removed in current operating systems
Skip it when always, unless a legacy device leaves you no alternative — and then assume the traffic is readable.
The question behind the question
Almost every “best VPN protocol” article ranks four protocols on a chart and declares a winner. That framing is wrong in a specific way: the protocols are not competing on one axis. WireGuard is faster than OpenVPN and always will be. OpenVPN gets through a hotel network that drops UDP and WireGuard does not. Neither fact beats the other, because they answer different questions.
So the recommender above leads with the network you are on, and treats everything else as an adjustment. In practice one distinction decides most of it: is anything actively blocking VPN traffic where you are? If no, take the fast one and stop thinking about it. If yes, no amount of speed helps, because the connection will not establish.
What the four options are actually for
WireGuard — the default for most people
Small, modern and fast. Roughly 4,000 lines of code against OpenVPN’s hundreds of thousands, which matters because it is a codebase a person can actually audit. It does not negotiate a cipher suite; it uses one set of modern primitives and has no downgrade path, which removes an entire category of attack.
Its weakness is that it is unmistakable. The handshake has a fixed structure and no cover story, so a network doing deep-packet inspection identifies it immediately. WireGuard was never trying to be invisible — it was trying to be fast and correct, and it is.
OpenVPN — the one that still gets through
Slower, older, and enormously more widely deployed. The reason to keep it available is TCP mode on port 443, which lets it cross networks that permit web traffic and nothing else — hotels and conference Wi-Fi, mostly.
Worth understanding why that mode is a last resort rather than a default. TCP guarantees delivery by retransmitting lost packets. Run TCP inside TCP and both layers retransmit the same loss, each making the other’s congestion control worse. On a clean link you barely notice; on a lossy one throughput collapses. It is called TCP meltdown, and it is why UDP is the right choice whenever UDP is available.
IKEv2/IPsec — the one for phones
Its distinguishing feature is MOBIKE, which lets a tunnel survive the client changing network address. Walk out of a building, drop off Wi-Fi onto mobile data, and the session continues rather than dropping and rebuilding. On iOS it is also native, so nothing third-party sits in the path.
It uses fixed, well-known ports — UDP 500 and 4500 — which are among the first things a restrictive network blocks. Excellent where it works; no help at all where it does not.
Obfuscated protocols — the one for filtered networks
A category, not a product. What these designs have in common is a goal: that traffic analysis should find nothing worth acting on. The better ones also survive active probing, where a censor connects to a server it suspects and watches how it responds — a server that answers strangely under probing is identified regardless of how its traffic looks.
It is important to be clear that this is not stronger encryption. The confidentiality is the same. What obfuscation buys is that the connection does not announce itself, and it costs speed to buy it.
What protocol choice does not fix
Three things people expect from a protocol decision that no protocol decision provides:
- What your provider can see. Every one of these terminates at the provider’s server. That is a question about the company’s logging policy and audits, not the cryptography.
- Your browser fingerprint. Unchanged by any of them — it comes from your device, not your connection. See what yours looks like.
- DNS leaking around the tunnel. A configuration failure, not a protocol property, and it happens with all of them. Worth testing rather than assuming.
The setting that genuinely matters more than protocol choice, and gets discussed far less, is the kill switch. A tunnel that drops without one puts you back on the bare network mid-session, silently, at exactly the moment you are least likely to check.
Protocol questions people actually have
Which VPN protocol is the fastest?
WireGuard, in essentially every published comparison and on essentially every device class. It uses fixed modern cryptography rather than negotiating a suite, runs in kernel space where the platform allows it, and carries far less per-packet overhead than OpenVPN. On a fast connection the difference is the gap between a VPN you notice and one you do not. The caveat is that raw protocol speed stops being the limiting factor once your server is far away — at that point the distance is your bottleneck, not the protocol.
Is WireGuard less private than OpenVPN?
The criticism is real but narrow, and it is about deployment rather than the protocol. WireGuard assigns each client a fixed internal IP address and keeps it for the session, which a provider could use to link activity to an account in a way OpenVPN's dynamic assignment does not. It is a solved problem — providers handle it with double-NAT or by rotating assignments — but it is worth knowing that it has to be solved rather than assumed. Cryptographically, WireGuard is the more modern and more reviewable of the two.
Do I need an obfuscated or stealth protocol?
Probably not, and it is worth being clear about that because it is the option most often oversold. Obfuscation exists to solve one problem: a network that identifies VPN traffic and blocks it. If your connection works, you do not have that problem, and the camouflage costs you speed for nothing. Where you do need it — national filtering, or a workplace or campus that actively cuts tunnels — nothing else reliably works, because the issue is not that conventional protocols are weak but that they are recognisable.
Should I ever use PPTP?
No. Its encryption has been practically breakable since 2012, and the attack is a service you can rent rather than a research result. Modern versions of Windows, macOS and iOS have removed or disabled it. If a piece of equipment offers you nothing else, treat that connection as unencrypted and plan accordingly.
Does the protocol change what my VPN provider can see?
Not meaningfully. Every protocol here terminates the tunnel at the provider's server, which means the provider is in a position to see your traffic as it leaves regardless of which one you picked. What determines your exposure there is the provider's logging policy and whether it has been independently audited — a question about the company, not the cryptography. Protocol choice decides what the networks between you and the server can tell, which is a different question.
Why does my VPN app not let me pick a protocol?
Often because there is only one worth offering, and sometimes because the app switches automatically based on what the network allows. Automatic selection is usually the right default — it will fall back from a fast protocol to one that survives a restrictive network without you having to diagnose the failure. It is worth being able to override it, though, for the case where you know your network filters and would rather start with the option that works than wait for two failures first.
Both answers exist in the same app.
Aurora is the fast default and what the recommender picks for most people. MIRAGE is the option for networks that filter, and it is slower on purpose — camouflage costs round trips. You do not choose once and live with it: switch in the app when a network stops cooperating, and switch back when it does.
Last reviewed . Found something out of date? Tell us.