Let $K,X$ be sets and let $F:K\times X\rightarrow X$ be a function. For each $k\in K$, let $f_{k}:X\rightarrow X$ be the function where $f_{k}(x)=F(k,x)$ whenever $k\in K,x\in X$. Assume that each $f_{k}$ is a bijection.
Suppose that $F$ is the round function for some cryptographic function such as AES-128 or some cryptographic function.
If $F$ is a cryptographic function, then I do not expect for $\{f_{k}\mid k\in K\}$ to generate the full symmetric group $S_{X}$, but I would expect for $\{f_{k}\mid k\in K\}$ to generate the alternating group $A_{X}$ (let me know if there are any real-world examples where $f_{k}$ is an odd permutation). Has there been cases in cryptography where it was rigorously proven that $\{f_{k}\mid k\in K\}$ generates or does not generate the alternating group $A_{X}$? For example, if $F$ is the round function for AES-128 or DES, then does $\{f_{k}|k\in K\}$ generate the alternating group $A_{X}$?
I am mainly interested in the case where the functions $f_{k}$ are the round functions because this case will probably be easier to analyze and because if the functions $f_{k}$ are the round functions, then it is more likely that $\{f_{k}\mid k\in K\}$ generates the alternating group.
This problem may be intractible in most cases, but there may be cases where one can show that $\{f_{k}\mid k\in K\}$ generates the alternating group $A_{X}$ such as outdated or insecure cryptography or when the cryptography has a special form that makes it easier to analyze (such as Feistel ciphers) or even cryptographic algorithms that are designed for testing.
We say that a subgroup $G$ of the permutation group $S_{X}$ is $n$-transitive if whenever $x_{1},\dots,x_{n}$ are distinct elements in $X$ and $y_{1},\dots,y_{n}$ are distinct elements in $X$, then there is some $g\in G$ with $g(x_{i})=y_{i}$ whenever $1\leq i\leq n$.
Theorem: Suppose that $X$ is finite and $|X|>24$. If $G$ is a $4$-transitive subgroup of $S_{X}$, then either $G=S_{X}$ or $G=A_{X}$.
The above theorem may make it easier to prove that $G=A_{X}$.
If $\{f_{k}\mid k\in K\}$ does not generate the alternating group $A_{X}$, then I would reject any block cipher with round function $F$ as being horribly insecure since either $|X|\leq 24$ which is too small for any block cipher or the group generated by $\{f_{k}\mid k\in K\}$ is not 4-transitive.
However, if it is easy or tractible to prove that $\{f_{k}\mid k\in K\}$ generates the alternating group $A_{X}$, then the function $F$ may be too well behaved for cryptographic purposes.
