We put a logger between a laptop and a test website, right where a hotel’s Wi-Fi would sit.
Without a VPN, it read our order number and session cookie word for word. With a WireGuard tunnel up, the same page load became 13 sealed packets and not one readable letter.
That’s how a VPN works, seen from the outside. But the logger wasn’t the only thing reading our traffic.
One line in our VPN server’s log shows the trade you make the moment you tap Connect.
How does a VPN work, in plain English?
A VPN encrypts everything your device sends and wraps it in a new packet addressed to the VPN server. The server unwraps it, sends your request on to the website, and seals the reply for the trip back.
Picture a letter inside a locked envelope. The postman reads only the outside address, and that address belongs to the VPN company, not the shop.
So what does a VPN do for you? It changes what three different parties can learn:
- Your network. The Wi-Fi owner and your ISP see sealed packets going to one server, instead of every site you open.
- The websites you visit. They see the VPN server’s IP address instead of the one your ISP gave you.
- The VPN company. It can now read whatever your ISP could read before.
Most explainers stop after the first two. The third decides whether a VPN is worth paying for at all, and we’ll come back to it.

So what does a sealed packet look like to the people in the middle? We built a lab to find out.
What can the Wi-Fi read? We logged every packet to find out
Our lab runs on a single Mac. WireGuard’s own open-source code builds a real tunnel between two programs. One plays your laptop; the other plays the VPN server.
Between them sits a relay that passes every packet along and logs what it can read. That’s the spot a café’s Wi-Fi or your ISP would occupy.
The website, its DNS server and the account details are test copies. The name shop.example is reserved for testing, and the session cookie is fake.
What we didn’t test: a commercial VPN app, a trip across the real internet, and speed. For overhead numbers, see our VPN tunnel test.
No VPN: the logger read everything
First, no tunnel. The laptop looked up shop.example, then asked for an account page over plain HTTP:
[ 0.50s] DNS question, plain text: "shop.example" (30 bytes, UDP port 53 in real life)
[ 0.50s] DNS answer, plain text: shop.example -> 127.0.0.1
[ 0.51s] laptop -> website, 137 bytes, readable:
GET /account?order=1042 HTTP/1.1
Host: shop.example
User-Agent: DroidCrunch-lab/1.0
Cookie: session=demo-7f3a9c
Connection: close
[ 0.51s] website -> laptop, 138 bytes, readable: HTTP/1.0 200 OK
[ 0.51s] website -> laptop, 39 bytes, readable: Order #1042: shipped. Arriving Friday.
It’s all there: the site, the page, the order number and the cookie that keeps you signed in. Whoever runs the network could copy that cookie and log in as you.
HTTPS: the page vanished, the name didn’t
Most sites use HTTPS now, so we repeated the request over HTTPS. The page and the cookie disappeared into encrypted records:
[ 0.82s] DNS question, plain text: "shop.example" (30 bytes, UDP port 53 in real life)
[ 0.82s] DNS answer, plain text: shop.example -> 127.0.0.1
[ 0.83s] TLS hello, laptop -> website: server name in plain text: "shop.example" (1517 bytes)
[ 0.83s] encrypted TLS records, website -> laptop: 2456 bytes | readable text: none
[ 0.83s] encrypted TLS records, laptop -> website: 80 bytes | readable text: none
[ 0.83s] encrypted TLS records, laptop -> website: 159 bytes | readable text: none
[ 0.83s] encrypted TLS records, website -> laptop: 255 bytes | readable text: none
[ 0.83s] encrypted TLS records, website -> laptop: 255 bytes | readable text: none
[ 0.83s] encrypted TLS records, website -> laptop: 221 bytes | readable text: none
But shop.example still showed up twice. Once in the DNS lookup, and once in the opening hello, where your browser names the site it wants.
So HTTPS hides what you read, not where you go. A newer feature, Encrypted Client Hello, can hide that hello. It only works when your browser and the site both support it.
With the VPN on: 13 packets, nothing readable
Then we switched the tunnel on and loaded the same page. The top pane is the VPN server’s log. The bottom pane is our relay’s view of the same page load.

Our relay found no readable text in any packet. The site name, the page and the cookie all travelled inside the encryption.
It still learned a few things. It saw which server the laptop talked to, when, and roughly how much data moved.
It could even tell this was a VPN. A WireGuard hello is always 148 bytes, and the first byte of every packet names its type. Our relay labelled the handshake without any keys, which is how some networks spot and block VPNs.
Real-world tip: if a network blocks your VPN, look for an obfuscation or stealth mode in the app’s settings. It disguises the tunnel as ordinary encrypted web traffic.
What happens to one request, step by step?
When you tap Connect, your app and the VPN server agree on fresh keys. From then on, every packet travels sealed between them.
One request through a VPN, in 6 steps
What our lab's WireGuard tunnel did, in order
-
Step 1 of 6
Your app and the server swap keys
The laptop sends a 148-byte hello carrying a fresh public key, and the server answers with 92 bytes. That one round trip sets up the keys for the whole session.
-
Step 2 of 6
Your device seals every packet
Each packet is encrypted with ChaCha20-Poly1305, given a counter and wrapped in a packet addressed to the VPN server. The counter stops replays, and the encryption stops tampering.
-
Step 3 of 6
Even your DNS lookup goes inside
Our laptop’s lookup for shop.example travelled sealed, and the VPN server’s own resolver answered it. If your lookups skip the tunnel, the site name leaks in plain text.
-
Step 4 of 6
The Wi-Fi and your ISP carry the envelope
They see your IP, the server’s IP, the packet sizes and the timing. Our relay found no readable text in any of the 37 packets it logged.
-
Step 5 of 6
The VPN server opens it and passes it on
The server removes the wrapper, reads the original request and sends it to the website from its own IP address. The reply comes back the same way, sealed for the trip down the tunnel.
-
Step 6 of 6
The keys renew themselves
We used the 25-second keepalive WireGuard’s docs suggest, so our idle laptop sent a 32-byte packet every 25 seconds. Its first request after the two-minute mark triggered a fresh handshake. Fresh keys keep older traffic locked even if a later key leaks.
Here’s the quiet part of our relay log, after the first page had loaded:
[ 12.55s] laptop -> VPN 32 bytes keepalive (empty, encrypted) readable text: none
[ 37.56s] laptop -> VPN 32 bytes keepalive (empty, encrypted) readable text: none
[ 62.56s] laptop -> VPN 32 bytes keepalive (empty, encrypted) readable text: none
[ 87.56s] laptop -> VPN 32 bytes keepalive (empty, encrypted) readable text: none
[ 112.56s] laptop -> VPN 32 bytes keepalive (empty, encrypted) readable text: none
[ 132.56s] laptop -> VPN 96 bytes encrypted data readable text: none
[ 132.56s] laptop -> VPN 148 bytes handshake: hello + key readable text: none
[ 132.56s] VPN -> laptop 112 bytes encrypted data readable text: none
[ 132.56s] VPN -> laptop 92 bytes handshake: reply readable text: none
Five empty keepalives, then a brand-new handshake the moment the laptop asked for the page again. Nobody pressed anything; the tunnel renewed itself.
Two of those steps carry most of the risk. If your DNS lookup skips the tunnel in step 3, the site name leaks in plain text. Our VPN tunnel guide caught exactly that on a laptop still using the Wi-Fi router’s DNS.
Step 5 is the other one. Hold that thought; it’s the trade from the intro.
One more failure sits outside the six steps: the tunnel dropping. A kill switch stops your device falling back to the open network. Our kill switch test shows what leaks without one.
Which protocol builds the tunnel, and does it matter?
The protocol is the rulebook your app follows to build the tunnel. Most apps give you some mix of three: WireGuard, OpenVPN and IKEv2.
- WireGuard is the newest and leanest. It uses one fixed set of ciphers and connects in a single round trip. That’s the 148-byte hello and 92-byte reply in our log.
- OpenVPN is older and heavier. Its trick is running over TCP port 443, the port HTTPS uses, so strict firewalls tend to let it through.
- IKEv2/IPsec reconnects quickly when your phone hops between Wi-Fi and mobile data. Most phones and laptops have it built in, and our IPsec guide covers how it works.

Some apps rename or rebuild these. NordVPN’s NordLynx runs on WireGuard, and ExpressVPN uses its own protocol, Lightway.
Our take: leave your app on WireGuard, or its version of it, unless a network blocks you. Then switch to OpenVPN over TCP. In our WireGuard vs OpenVPN test, WireGuard moved about three times more data per second.
Who sees what once you’re connected?
With a VPN on, your Wi-Fi and ISP see only encrypted traffic to one server. Websites see the server’s IP address. The VPN company sees which sites you visit.
Remember the line in our VPN server’s log from the intro? Here it is, with the two lines around it:
[ 2.01s] VPN server: decrypted a DNS question: shop.example (A)
[ 2.02s] VPN server: decrypted a request: GET http://shop.example/account?order=1042 Cookie: session=demo-7f3a9c
[ 2.02s] VPN server: passed it to the website, got 39 bytes back, sending them down the tunnel encrypted
The relay saw nothing, but the server saw the site, the page and the cookie. It has to open your packets to send them on. So it can read whatever travels unencrypted inside them.
That’s the trade. A VPN doesn’t remove the party that can read your traffic; it swaps your ISP for the VPN company.
On an HTTPS site, the server sees less: the site name and the time, not the page or your password. That’s still a list of every site you visit, so pick a company you’d trust with it.

Warning: free VPNs still pay for their servers, usually with ads, data caps or upsells. Read the privacy policy before you send your traffic through one.
Does a VPN make you anonymous?
No. A VPN hides your IP address and browsing from the network. Websites can still recognise you in other ways:
- Logins and cookies. Sign in to Google and Google knows it’s you, whatever IP you arrive from. The cookie in our lab reached the website untouched.
- Browser fingerprints. Your screen size, fonts and browser settings form a pattern that sites can follow across IP addresses.
- Leaks around the tunnel. A DNS lookup or a WebRTC call that skips the VPN can expose your real address. Our leak test guide shows how to check.
- Phishing and malware. A VPN protects the road, not the destination. A fake login page works fine through a tunnel, and two-factor authentication helps more there.
Need anonymity, not privacy? Tor sends you through three volunteer relays, so no single relay knows both who you are and where you’re going. Our proxy vs VPN guide explains why a proxy won’t get you there either.
Does a VPN slow your internet down?
A little. WireGuard adds 60 bytes to each packet on IPv4. The extra distance to the server usually costs more than that.
Those 60 bytes are 32 for the WireGuard wrapper and 28 for the new outer address. Our tunnel test measured the total: a 5 MiB download took 4.5% more bytes through the tunnel.
Distance matters more. Every packet now detours through the VPN server, so pick one close to you unless you need another country.
Still slow? Try WireGuard first, then a less crowded server nearby. Speed rarely comes down to the encryption itself.
What should you check before you trust a VPN?
Check four things: an audited no-logs policy, a kill switch, DNS that stays in the tunnel and WireGuard support.
- An audited no-logs policy. Our lab’s server saw everything, so you want proof the company keeps none of it. That proof should come from an outside auditor, not the marketing page.
- A kill switch that’s on by default. Without one, our kill switch test showed the real IP in 10 of 30 checks during a server switch.
- DNS inside the tunnel. The app should answer lookups with its own resolver, like our lab’s server did. Run a leak test on day one.
- WireGuard, or the app’s version of it. It’s the fastest of the three, and its small codebase is easier to audit.
Ownership and location matter too. A company in a country that can order it to keep logs can’t promise much.
On a budget? Our free VPN list is a starting point; run the same four checks before you rely on one.
Affiliate note: we earn a commission if you sign up through the links below, at no extra cost to you. They don’t change what we recommend.
- NordVPN: NordLynx, its WireGuard-based protocol, is the default.
- Surfshark: plain WireGuard, OpenVPN and IKEv2, with no cap on devices.
- ExpressVPN: its own Lightway protocol, if you never want to touch a setting.
We haven’t run these apps through this lab. Read our NordVPN, Surfshark and ExpressVPN reviews, and run the four checks yourself on day one.
Frequently asked questions
Can my ISP see what I do when I use a VPN?
No. Your ISP sees that you’re connected to a VPN server, plus how much data you send and when. It doesn’t see the sites or pages. It can usually tell you’re using a VPN, though, because the server’s address and the packet pattern give it away.
Is a VPN safe to use on public Wi-Fi?
Yes, and it’s one of the best reasons to have one. On a hotspot you don’t control, a VPN stops the owner from reading or changing your traffic. It also protects you from fake hotspots that copy a real network’s name. Connect before you open any app, because phones start syncing the moment they join.
Can a VPN company see my passwords?
Not on HTTPS sites, which covers almost every login page today. The provider sees which sites you visit and when, but your password stays inside the site’s own encryption. On a plain HTTP page, the VPN server could read it, the same way ours read the test cookie.
Do I need a VPN if every site already uses HTTPS?
It depends on who you’re hiding from. HTTPS protects what’s on the page. A VPN also hides which sites you visit from your Wi-Fi and ISP. It swaps your IP address for the server’s, too. If neither matters to you, HTTPS covers the basics.
Does a VPN hide my IP address from websites?
Yes. Websites see the VPN server’s IP address, which also changes the country they think you’re in. Your real address can still slip out through a WebRTC request or a DNS lookup outside the tunnel. Test for leaks once after you connect.
Can the police track someone who uses a VPN?
They can’t read the traffic on its way, but they can ask the VPN company for its records. What it hands over depends on what it keeps. That’s why an independently audited no-logs policy matters more than any marketing claim.
Does a VPN use more mobile data?
Slightly. Every packet carries extra headers, which added about 4.5% to a download in our tunnel test. Idle connections add a trickle too: our lab’s laptop sent a 32-byte keepalive every 25 seconds. On most mobile plans, you won’t notice either.
Is it legal to use a VPN?
In most countries, yes, including the US, the UK and India. China and Russia allow only government-approved VPNs. India requires VPN providers to keep customer records for five years. Several moved their servers out of the country in response.
So, how does a VPN work, and should you use one?
Here’s how a VPN works in one line: it seals your traffic and sends it through a server you choose. Your network sees an envelope, and websites see the server.
Use one on public Wi-Fi, and whenever you’d rather your ISP didn’t see every site you open. Pick WireGuard, keep the kill switch on and check for leaks.
Don’t count on it for anonymity, though. That’s Tor’s job.
One warning: the VPN server now reads what your ISP used to. An audited no-logs policy matters more than any speed chart.
Already paying for a VPN? Spend five minutes on the test in our kill switch guide today.



