Traditionally, due to the way the network is structured, each input has a set of weights, that are connected to more inputs. If the inputs switch, the output will too.
Approach 1
However, you can build a network that approaches this behaviour. In your training set, use batch learning and for each training sample, give all possible permutations to the network such that it learns to be permutation invariant. This will never be exactly invariant, it just might be close.
Approach 2
Another way to do this is to have the weights replicated for all inputs. For example, let's assume you have 3 inputs ($i_0, i_1, i_2$), and the next/first hidden layer has 2 nodes ($h_{10}, h_{11}$) and activation function $\phi$. Assuming a fully connected layer, you have 2 weights $w_0$ and $w_1$. The hidden layer's nodes $h_{10}$ and $h_{11}$ are given, respectively, by
Thus giving you a hidden layer whose values are permutation invariant from the input. From now on, you can learn and build the rest of the network as you see fit. This is an approach derived from convolutional layers.
Approach 3
Finally, you can use a dimension reduction to achieve this. I've published this in Exploring Communication Protocols and Centralized Critics in Multi-Agent Deep Learning, on Integrated Computer-Aided Engineering. This approach uses convolutional architectures to efficiently achieve permutation invariance.