39 lines
2.2 KiB
Markdown
39 lines
2.2 KiB
Markdown
|
# Network wall
|
|||
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
|
> the best security = no network = no incoming tcp/ip from internet to containers
|
|||
|
|
|
|||
|
|
This is done via sockets.
|
|||
|
|
|
|||
|
|
- TCP router client opens up socket to TCP router server, residing on the web gateway.
|
|||
|
|
- When http arrives on this tcp router server, payload van http is brought back over socket to tcp router client.
|
|||
|
|
- TCP router client sends http request that is made to server residing in the container.
|
|||
|
|
- No TCP comes from outside world to the container, creating the most secure connection possible.
|
|||
|
|
- The TCP router client opens the socket, TCP router server that received http request throws it on the socket.
|
|||
|
|
- On the socket there is only data that comes in, which is replayed. TCP router client does a https request.
|
|||
|
|
|
|||
|
|
This mechanism is now implemented for https, but in the future also other protocols such as sql, redis, http … can be supported.
|
|||
|
|
|
|||
|
|
The end result is that only data goes over the network.
|
|||
|
|
If container can no longer go to local tcp stack but only to make outgoing connection to the gateway, then there is no longer tcp coming’s in from outside.
|
|||
|
|
|
|||
|
|
This is what we call the 'Network wall'.
|
|||
|
|
As a consequence, no tcp/ip is coming in AT ALL, making the full set-up reach unprecedented security.
|
|||
|
|
|
|||
|
|
|
|||
|
|
## More detailed explanation
|
|||
|
|
|
|||
|
|
- Containers are behind NAT. We don’t allow traffic coming in.
|
|||
|
|
- All connection needs to come from container to the outside world. = very neat as network security.
|
|||
|
|
- Connection needs to be outwards, secures against DDOS and other hacking etc, nothing to connect to.
|
|||
|
|
- How to access it ? Drive traffic inside the container: proxy or load balancer which is exposed publicly, rest of the traffic in private network, not accessible.
|
|||
|
|
- So you can limit the number of containers that are really accessible from the outside world.
|
|||
|
|
- You don’t have to worry about ‘how to secure my DB’ as my DB is not exposed, only accessible in Wireguard private network.
|
|||
|
|
- In containers you can specify to have specific IPv6 address, so deploy reverse proxy in container which has public access, = entry point in the network, deploy reverse tcp connection (=tcp router client), connects to the gateways and allows incoming connection.
|
|||
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
|
|