
Core takeaway
- Hidden by default: your server is unreachable from outside until you change that, and the wrong choice sits quietly open until a scanner finds it, which usually takes hours, not weeks.
- Four ways in: port forwarding, a VPN (WireGuard or Tailscale), a tunnel, or a VPS as a relay. The guide shows what each one exposes.
- Check your work: test what the internet can see, and know what to do if you think you were exposed.
- Rule: open as little as possible, and know exactly what you opened.
Disclosure: con1ainer.com participates in affiliate programs. This article contains no affiliate links. See the Affiliate & Commercial Disclosure.
A home server works perfectly from your couch. The trouble starts when you want it from a hotel, an office, or your phone on mobile data.
That is the moment you decide how much of your home network the internet gets to see. The wrong choice does not break anything visibly. It sits quietly open until a scanner finds it, which usually takes hours, not weeks.
This guide explains why your server is unreachable from outside by default, the four ways to change that, what each one exposes, and how to set up the safest option step by step. Every term is explained as it appears.
It builds on Best Mini PCs for a Home Server: What to Buy, and it assumes your server is running and you have followed the basics in How to Secure a VPS: Essential Setup Checklist, which apply at home too.
Why Your Server Is Hidden by Default
Your internet provider gives your home one public IP address, the address the rest of the internet sees. Inside the house, every device has a private IP address, usually starting with 192.168. or 10., which means nothing outside your network.
Your router connects the two using NAT (network address translation). When your laptop requests a web page, the router sends the request out under the public address, remembers who asked, and hands the reply back to the laptop.
The important part is the other direction. If something on the internet contacts your public address unprompted, the router has no record of anyone asking, and no idea which device inside should receive it. It drops the traffic. That default is protection, and every method in this guide either works around it or leaves it alone.
A few related terms:
| Term | What it means |
|---|---|
| Port | A numbered entrance on a computer. Web traffic uses 80 and 443. SSH uses 22. Each program listens on its own |
| Dynamic IP | A public address your provider may change without warning, which breaks anything that points to it |
| DDNS | Dynamic DNS. A name such as home.example.com that a small updater keeps pointed at your changing public address. Many routers have it built in |
| CGNAT | Carrier-grade NAT. Your provider shares one public address among many customers, so you never get an address of your own |
Check whether you are behind CGNAT
CGNAT matters because it makes port forwarding impossible. To check, find your router’s WAN address, which is the one it uses to face your provider, in its status page. Then, from any computer at home, find the address the world sees:
curl -4 ifconfig.me
If the two match, you have your own public address. If the router shows a different address, especially one starting 100.64 through 100.127 or a private range, you are behind CGNAT. The methods that do not need an open port, a VPN such as Tailscale, a tunnel, or a VPS relay, are the ones that will work for you.
NAT is not a firewall, especially with IPv6
IPv6 is the newer address system. Many connections now provide it, and with IPv6 your devices can have public addresses and no NAT in between. Protection then comes entirely from your router’s firewall, which should block unsolicited incoming traffic by default. Check that your router’s IPv6 firewall is on and set to block inbound connections, and that no setting allows everything through.
Four Ways to Reach Your Server
| Method | How it works | What the internet can see | Works behind CGNAT | Effort | Best for |
|---|---|---|---|---|---|
| Port forwarding | The router passes chosen incoming ports to your server | Your home IP, and every forwarded service | No | Low | Public services you are ready to defend |
| VPN (WireGuard) | An encrypted tunnel from your device to your server or router | One quiet UDP port, which does not answer strangers | No, unless combined with a relay | Medium | Private access to your own tools |
| Mesh VPN (Tailscale) | Your devices join a private network and find each other automatically | Nothing, since no ports open | Yes | Lowest | Easy private access, even behind CGNAT |
| Tunnel (Cloudflare Tunnel and similar) | Your server makes an outbound connection to a provider, who passes visitors in | Nothing from your home. The provider’s address appears | Yes | Medium | Sharing a website without opening a port |
| VPS relay | Your home server connects to your own VPS over WireGuard, and the VPS faces the public | The VPS, not your home | Yes | Higher | Public services with your home IP hidden and no third-party provider |
The next sections cover each in turn, starting with the one to avoid unless you have a reason.
Method 1: Port Forwarding
Port forwarding is a rule in your router: send anything arriving on public port X to this internal device on port Y. It is the traditional method, it is simple, and it is the riskiest.
The risks are concrete:
- Anything forwarded is reachable by everyone. Automated scanners probe every public address for open ports around the clock. A newly opened port is usually found within hours.
- Every forwarded service is attack surface. A flaw in that program is a way into your home network.
- Your home address is exposed, and so is, roughly, where you live.
- You become the one patching. An unpatched service is the usual way home servers get compromised.
If you do forward, follow these rules.
| Rule | Why |
|---|---|
| Forward only 443 (HTTPS), and only to a reverse proxy | One well-defended entrance, not a dozen |
| Never forward databases, file shares (SMB), remote desktop (RDP), the Docker API, or admin dashboards | These are the most commonly attacked services and none should face the internet |
| Never forward SSH with password login | Password logins are guessed constantly. If you must, use keys only and no root login |
| Give the server a fixed internal address | Otherwise your rule points at the wrong device after a reboot |
| Use HTTPS with a real certificate | Without it, logins cross the internet in plain text |
| Require strong logins and two-factor authentication | The login page is the front door |
| Turn off UPnP on the router | UPnP lets programs open ports without asking you, so you lose track of what is exposed |
| Keep everything updated and monitor the logs | An exposed service stays safe only while it is maintained |
A reverse proxy is a web server in front of your applications, as described in Docker on a VPS: A Practical Starting Point. The setup is the same at home as on a VPS.
If you read that table and think it is a lot of work, that is the point. For private tools used only by you, one of the next methods is safer and easier.
Method 2: A VPN
A VPN (virtual private network) is an encrypted tunnel between two devices. When your phone connects to a VPN that ends at your home, your phone behaves as though it were sitting on your home network, and your server is simply there.
Nothing is published to the internet. Instead of exposing each service, you expose one quiet entrance, and only devices holding the right key can use it.
WireGuard
WireGuard is a modern, fast, small VPN that is built into the Linux kernel. Each device has a pair of keys, a private key it keeps secret and a public key it shares. A device is allowed in only if its public key is listed. WireGuard also does not reply to packets that fail authentication, so a port scan sees nothing at all.
This setup lets your phone or laptop reach the server itself. It needs one open UDP port, so it requires your own public address and does not work behind CGNAT.
1. Install it and create the server’s keys:
sudo apt install wireguard
sudo -i
umask 077
wg genkey | tee /etc/wireguard/server.key | wg pubkey > /etc/wireguard/server.pub
exit
The umask 077 line makes new files readable by their owner only, which matters for a private key.
2. Create a key pair for each device (a phone, in this example), in a private folder:
mkdir -p ~/wg-clients && cd ~/wg-clients
umask 077
wg genkey | tee phone.key | wg pubkey > phone.pub
3. Write the server configuration. Open the file:
sudo nano /etc/wireguard/wg0.conf
[Interface]
Address = 10.8.0.1/24
ListenPort = 51820
PrivateKey = PASTE_SERVER_PRIVATE_KEY_HERE
[Peer]
# phone
PublicKey = PASTE_PHONE_PUBLIC_KEY_HERE
AllowedIPs = 10.8.0.2/32
The private key is the content of /etc/wireguard/server.key, and the public key you paste in comes from phone.pub. 10.8.0.1 is the server’s address inside the tunnel, a private network that exists only for your VPN. AllowedIPs in the server’s file says which address this device is allowed to use.
4. Start it, and open the firewall:
sudo chmod 600 /etc/wireguard/wg0.conf
sudo systemctl enable --now wg-quick@wg0
sudo ufw allow 51820/udp
sudo wg show
5. Forward one port on the router. Forward UDP port 51820 to the server’s fixed internal address. That is the only opening in your network.
6. Write the phone’s configuration:
[Interface]
PrivateKey = PASTE_PHONE_PRIVATE_KEY_HERE
Address = 10.8.0.2/24
[Peer]
PublicKey = PASTE_SERVER_PUBLIC_KEY_HERE
Endpoint = home.example.com:51820
AllowedIPs = 10.8.0.0/24
PersistentKeepalive = 25
Endpoint is your DDNS name and the port. AllowedIPs lists the destinations that go through the tunnel, here only the VPN’s own network, so the rest of your phone’s traffic is not affected. PersistentKeepalive sends a small message every 25 seconds to keep the connection alive through routers.
Import it into the official WireGuard app. The simplest way is a QR code:
sudo apt install qrencode
qrencode -t ansiutf8 < phone.conf
Scan the code in the app. Treat the QR code like a password, because it contains the phone’s private key. Delete the file when finished.
Switch the phone to mobile data, turn on the tunnel, and open an address such as http://10.8.0.1:8080. If a page loads, the setup works.
Tailscale
Tailscale is built on WireGuard but manages everything for you. You install it on each device and log in. Your devices join a private network and find each other automatically, even when both are behind CGNAT, without any router changes.
curl -fsSL https://tailscale.com/install.sh | sh
sudo tailscale up
The first line downloads and runs the installer from Tailscale’s own site. Running a script straight from the internet means trusting that site, so read the script first if you prefer. The second line prints a link; open it, log in, and the server joins your network. Install the app on your phone and log in with the same account. Each device then gets an address starting 100., and a name you can use instead.
The trade-offs:
- No open ports, and it works behind CGNAT. Both are major advantages.
- It depends on a company’s coordination service to introduce your devices to each other. The traffic itself is encrypted between your devices. Headscale is a self-hosted alternative if you want to avoid the dependency.
- Free plan limits and features change, so check them before committing.
For most people with a home server, Tailscale is the easiest safe answer, and WireGuard is the right one if you want no third party in the loop.
Method 3: A Tunnel
A tunnel service such as Cloudflare Tunnel reverses the usual direction. A small program on your server makes an outbound connection to the provider and keeps it open. When someone visits your address, they reach the provider, who passes the request down the existing connection to your server.
Because the connection starts from your side, you open no port, your home address stays hidden, and it works behind CGNAT. The setup, in outline, is: install the provider’s connector program, log in, create a tunnel, point a hostname at it, and tell it which local address to forward to, such as http://localhost:8080. The commands and dashboards change over time, so follow the provider’s current setup guide.
Two things to know before you build on one.
- The provider sees the traffic. For web traffic, the provider typically decrypts it on its own servers and re-encrypts it toward you. If the content is private, you are trusting them.
- Terms of service apply. Some providers restrict heavy use, such as streaming large amounts of video, through free tunnels. Read the current terms before building something on them.
Tunnels also let you add a login screen in front of the service, handled by the provider, so strangers never reach your application at all. That is a strong option when you want to share something with people who will not install a VPN app.
Method 4: A VPS as a Relay
If you already run a VPS, you can combine its advantages with your home hardware. Your home server makes a WireGuard connection out to the VPS. The VPS has a stable public address and does the facing. Visitors reach the VPS, and the VPS passes the traffic through the tunnel.
You get no open ports at home, a hidden home address, a stable address, no third-party tunnel provider, and it works behind CGNAT. The cost is a small VPS bill and more to set up.
The VPS is the WireGuard server, with address 10.8.0.1, and the home server connects to it as 10.8.0.2. On the home server’s configuration:
[Interface]
PrivateKey = PASTE_HOME_PRIVATE_KEY_HERE
Address = 10.8.0.2/24
[Peer]
PublicKey = PASTE_VPS_PUBLIC_KEY_HERE
Endpoint = vps.example.com:51820
AllowedIPs = 10.8.0.1/32
PersistentKeepalive = 25
On the VPS, add the home server as a peer with AllowedIPs = 10.8.0.2/32, and open UDP port 51820. Then add an Nginx server block on the VPS that passes requests to the home machine through the tunnel:
server {
server_name media.example.com;
location / {
proxy_pass http://10.8.0.2:8096;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
}
}
Issue the HTTPS certificate on the VPS with Certbot, exactly as in the other guides. On the home server, allow that port only from the tunnel:
sudo ufw allow in on wg0 to any port 8096
The nginx and certificate steps are in How to Move a Website to a VPS Without Downtime. Pick this method when you want a public service, a stable address, and no home exposure.
Which Method for Which Job?
| What you want | Best method | Why |
|---|---|---|
| Reach private tools, files, and dashboards yourself | Tailscale or WireGuard | Nothing public, and the least to defend |
| The same, but your provider uses CGNAT | Tailscale, or a VPS relay | No open port needed |
| Share a website or app with the public | Tunnel or VPS relay | Hides your home address, and the front door is someone else’s |
| Family members who will not install anything | Tunnel with a login screen in front | They open a normal link |
| SSH into the server while travelling | VPN, then SSH over the tunnel | SSH is never exposed |
| A public service with no third-party provider | VPS relay | You control every part |
| You have a reason to forward ports | Port forwarding, with the rules above | Accept the maintenance |
A rule of thumb: private things go behind a VPN, public things go behind a tunnel or relay, and port forwarding is the last resort.
Protections That Apply to All of Them
Whichever method you use, these matter.
- Keep the server updated, including containers. Exposed or not, old software is how problems start.
- Use strong, unique passwords and two-factor authentication on every login that can be reached.
- Separate risky devices. If your router supports VLANs or a guest network, put cheap smart-home gadgets on a different network from the server.
- Turn off UPnP and unused router features.
- Monitor it. Watch the logs and add an outside uptime check, as described in How to Monitor a VPS: Uptime, Resources, and Alerts. A home internet outage then reaches your phone.
- Back up off-site. A compromised or failed machine at home needs a copy somewhere else. See VPS Backup Strategy: How to Back Up a VPS and Actually Restore It
- Add a UPS, so a power cut does not corrupt the machine. See Best UPS for a Home Server: What to Buy and How to Size It
Test what the internet can see
Check your setup from outside your network, using mobile data or a VPS, rather than from your couch. From a VPS or another network, scan your own public address for open ports. Only ever scan addresses you own:
sudo apt install nmap
nmap -Pn -p 1-1024 home.example.com
-Pn skips the check that the host is up, which some routers do not answer, and -p 1-1024 scans the first 1024 ports. With a VPN-only setup, you should see only the WireGuard port if it is UDP, which this TCP scan will not show, and nothing else. Anything unexpected that shows as open is a rule you forgot, or one UPnP created.
Common Problems and What They Mean
| Symptom | Likely cause | Fix |
|---|---|---|
| Port forwarding rule has no effect | You are behind CGNAT, or the router has a second firewall layer | Check for CGNAT. Use Tailscale, a tunnel, or a VPS relay |
| The VPN connects, but a service does not load | The service is bound to 127.0.0.1, which only the server itself can reach | Bind it to the VPN address, such as 10.8.0.1:8080:80 in the Compose file |
| Docker fails to start a container bound to the VPN address | The WireGuard interface is not up yet when Docker starts | Make sure wg-quick@wg0 starts first, then restart the container |
| VPN worked yesterday and fails today | Your public IP changed and DDNS did not update | Check the DDNS record and updater |
| Connected VPN, but no traffic flows | Wrong key, or the wrong AllowedIPs on one side | Check each public key matches, and run sudo wg show for a recent handshake |
| Server unreachable after a power cut | The BIOS setting to restart after power loss is off | Enable Restore on AC Power Loss |
| Slow remote access | Your home upload speed is low | Check your plan’s upload speed, because that limits remote use |
| Certificate warnings | The certificate does not match the name you used | Use the DNS name the certificate was issued for |
A handshake is the moment two WireGuard devices confirm each other’s keys. sudo wg show lists the time of the latest handshake for each peer. If it is empty, the two sides are not agreeing, or traffic is not arriving.
If You Think You Were Exposed
Signs include unknown logins in your logs, unexpected processes, a sudden rise in traffic, or unfamiliar containers. If you see them:
- Disconnect the machine from the network.
- Change passwords and keys for everything the machine could reach, from another device.
- Do not trust the machine. Back up your data, wipe the drive, reinstall, and restore the data only.
- Find the way in. Review which ports were forwarded and which software was exposed.
- Rebuild with a safer method, such as a VPN.
Frequently Asked Questions
Is port forwarding always unsafe?
No. A well-maintained reverse proxy with HTTPS, strong logins, and two-factor authentication on port 443 is a reasonable setup, and many people run one. It is simply the method with the most to maintain, and the least forgiving of neglect.
Do I need a static IP address?
Not for a VPN or tunnel, and not even for port forwarding, if you use DDNS. A static IP is a convenience, and it often costs extra.
Is Tailscale safe?
Its traffic is encrypted between your devices, and it opens no ports on your network. The thing you are trusting is the company’s coordination service and your account’s login security. Turn on two-factor authentication for the account.
Can I just use my router’s built-in VPN?
Many routers include WireGuard or OpenVPN. It can be a simple option if you keep the router updated. Router firmware varies in quality and update frequency, so check that your model is still supported.
Which is better, WireGuard or Tailscale?
Tailscale is easier and works behind CGNAT. WireGuard has no third party and fewer features to learn. If you are unsure, start with Tailscale.
What about the apps on my phone?
Both WireGuard and Tailscale have official apps for iPhone and Android. Once connected, you open your usual apps and addresses.
Open as Little as Possible
The safest way to reach a home server is to publish nothing. Put private things behind a VPN, put public things behind a tunnel or a relay, and treat port forwarding as the exception that needs a reason.
Whichever you choose, test it from outside your network, keep everything updated, and back up somewhere that is not your house.
Open as little as possible, and know exactly what you opened.