$\DeclareMathOperator{\Char}{char}$$\DeclareMathOperator{\Aut}{Aut}$ Let $G$ be a group, $H$ a subgroup, such that $H \Char G$. If $\phi \in \Aut(H)$ is there an automorphism $\widehat{\phi} \in \Aut(G)$ such that $\widehat{\phi}|_H$ = $\phi$?
In words, if $\phi$ is an automorphism of $H$, is there necessarily an automorphism $\widehat{\phi}$ of $G$ where $\widehat{\phi}$ maps the elements in $H$ to $H$ in exactly the same way?
I am thinking this is false (consider the non-finite example of $\bar{Q}$ and $\mathbb{C}$), but would like a (several) counter example(s), preferably for the finite case, but I wouldn't mind another infinite case.
The reason why I am not satisfied with $\bar{Q}$ and $\mathbb{C}$ is that, according to some other answers on MO/Math.stack, the automorphisms of $\mathbb{C}$ depends on whether we are willing to use AC or not (as far as I understood those posts), so to me they are not really "concrete" counter-examples, (unless someone can convince me otherwise).