I will assume you are using the RANS Flow solver and that you have successfully completed the tutorial "Incompressible Flow Through a Pipe into a Box", which is similar in configuration to the simulation you have described.

There are no orientation options for boundary conditions (e.g., inlets or outlets). The faces are implicitly orientated correctly if you have a watertight volume. Are you sure the volume is watertight and the mesh is correct? If not review the comment "Diagnose geometry, face meshes, or volume mesh problems".

The default outlet evaluates the velocity upstream and imposes that velocity at the outlet. It doesn't care whether it is moving in or out of the flow volume. This is called a 'zero gradient' condition. So if flow is moving into the domain through the outlet it is because of the upstream conditions.

Assuming you have correctly assigned your boundary conditions and set your reference velocity equal to your inlet velocity, then a couple of explanations come to mind:

1. Have you converged your simulation as in the tutorial? You should aim to see your residuals drop at least 3 orders of magnitude, i.e., below 10^-3.
2. Is your outlet in a steady region of the flow, i.e., away from velocity gradients? It could be that the outlet is situated close to a recirculation zone, especially if there's a rapid geometry expansion upstream. If this is the case then try moving your outlet further downstream.