I need to perform attacks (SQL injection) on a server. I thought of using a proxy to be anonymous, but the topology is a LAN without Internet access. I have no idea of the protocols used in the topology.
Anonymity on a LAN is a bit of an odd question but here are some points to address which I feel may aid in your quest.
Firstly if you are on the LAN plugging in your laptop via an Ethernet port, I know where you are. Your physical being would easily be identifiable by anyone choosing to look.
For the sake of discussion let us assume a nice easy scenario, you are connecting to an open WiFi using a high gain antenna from a safe distance, this WiFi simply being a poorly secured access point to give access to the LAN via wireless means.
Change the MAC of your network interface
A mac Address is made up of 6 parts 00:D0:94:00:00:00 The first 3 parts make the vender ID (who is the creator of your device) the second part is the unique identifier of your device. While it could take a lot of effort to find out who has this MAC, it is just good house keeping (especially if you have used the MAC on the network for legitimate purposes)
Change your hostname
This is just another piece of the puzzle; and can leak information as to what device you are using. default host names can make it fairly obvious what you are using.
Block as much traffic as you can
Firewall off your device, you want as little information leakage as possible when remaining anonymous, lets say you accidentally a service running (e.g. a web browser). Once again its the puzzle as a whole, not each issue by its self.
I would recommend attaching an network card to a VM and running the attack from inside it, keeping your host segregated from the network.
So many things
You could use so many different ways, such as network implants, compromising other machines, stealing user credentials, using an onsite work station. the list is truly massive. For a more relevant answer to you, you would need to provide a lot more information of the network, your proximity, the size/type of organisation and much more than what you have provided
Disclaimer There is no guarantee in all of this but its a good foundation
If the written contract which asks you to perform penetration testing at this organization also includes to check if the end users follow the security policies correctly, then this might be a good opportunity to check if everyone locks their PC when they leave their office. Using an unattended PC of someone else to demonstrate a vulnerability can be an effective way to demonstrate the importance of IT security to the end-users. It is a way to teach them that the security rules do not just protect the company but also protect them personally from wrong accusations.
But the question is what kind of security vulnerability you are trying to prove here. If an intranet web application is vulnerable to SQL injections, then it's vulnerable from every workplace which can access it. If you want to prove that the way they do access logging is unreliable, then you don't need to do so by performing an SQL injection.
If using Linux, set your firewall to deny everything that's not allowed, spoof your hostname, change your MAC address. If you are using Windows then leave this place.
Or maybe try to do it from another client.