Let $f: \mathbb{R}^{m} \to \mathbb{R}$ positively homegeneous of degree $k \in \mathbb{N}$ and $f \in C^{k}$. Then there exists a $k$-linear form $\varphi$ in $\mathbb{R}^{m}$ such that $f(x) = \varphi(x,...,x)$ for all $x \in \mathbb{R}^{m}$.
I know that:
$$f(tx) = t^{k}f(x),$$
and
$$\sum \frac{\partial f}{\partial x_{i}}x_{i} = kf(x) \quad (\textit{Euler relation}),$$
If $\varphi$ is a $k$-linear form, so $$\varphi(x_{1},...,x_{k}) = Cx_{1}\cdot ... \cdot x_{k}.$$
But, I don't know how I can use this informations. I appreciate any hints.
Thus is not true in general. The simplest counterexample is the case $m=1$ and $f(x):=\vert x\vert^k$. For arbitrary $m$ you may do the same by using any norm on $\mathbb{R}^m$ instead of the absolute value.
Let now $f$ be in $C^k$ with $k\geq 1$. (Reaction on the remark below.)
Consider the function $\mathbb{R}^n\times[0,\infty)\ni (x,t)\mapsto f(t\cdot x):=h(x,t)$. Then, using the chainrule, $\frac{\partial h}{\partial x_{i}}(x,t)=\frac{\partial f}{\partial x_{i}}(t\cdot x) t$. Since $h(x,t)=t^k f(x)$ we also have $\frac{\partial h}{\partial x_{i}}(x,t)=t^k \frac{\partial f}{\partial x_{i}}(x)$. Thus $\frac{\partial f}{\partial x_{i}}(t\cdot x)=t^{k-1} \frac{\partial f}{\partial x_{i}}(x)$ for $t>0$ and by continuity also for $t=0$. So (induction) $\frac{\partial f}{\partial x_{i}}(x)=P_i(x)$ where $P_i$ is a homogeneous polynomial of degree $k-1$.
Let $g(t):=f(tx)$. Then $g'(t)=\sum_{i=1}^n\frac{\partial f}{\partial x_{i}}(tx) x_i =\sum_{i=0}^n P_i(tx)x_i=t^{k-1}\sum_{i=0}^n P_i(x)x_i=t^{k-1} P(x)$ where $P$ is a homogeneous polynomial of degree $k$.
Now $g(1)=f(x), g(0)=0$ and $f(x)=g(1)-g(0)=\int_0^1 g'(t)=P(x)\frac{t^k}{k}{\Large\vert}_0^1=\frac{P(x)}{k}$, a homogeneous polynomial of degree $k$.
Finally it is rather well-known that any homogeneous polynomial is the `diagonalization´ of a $k$-linear form. See here or the book on differential calculus by Henri Cartan.