I saw this question late but I feel my contribution is not late.
Note that for sure the identity is mapped to itself. Let $A_1=<(a,e)>$ and $A_2=<(e,a)>$. Observe that if the image of $(a,e)$ is fixed then automatically all other non-identity elements in $A_1$ (i.e.,$(a,e)^i$ with $2\leq i \leq p-1$) have their images determined. $(a,e)$ has $p^2-1$ choices as each non-identity element has order $p$. So all elements in $A_1$ have their images determined. Similarly, we note that $(e,a)$ has $p^2-p$ choices left and all other non-identity elements in $A_2$ have their images automatically fixed. Also note that by fixing images of $(a,e)$ and $(e,a)$ we have automatically fixed the image of an element of the form $(a_1,a_2)$, where $a_1,a_2\neq e$, as it is equal to $(a_1,e)(e,a_2)$. This means that $o(Aut(G))=(p^2-1)(p^2-p)$.
Hope this helps!