Let $\Pi$ denote the encryption scheme you just described, and $\mathcal{A}$ be the adversary trying to attack this scheme.
Define the experiment $\mathsf{PrivK}^{\mathsf{cca}}_{\mathcal{A}, \Pi}$ as follows:
- $\mathcal{A}$ has access to the encryption oracle $E_k(\cdot)$ and the decryption oracle $D_k(\cdot)$.
- $\mathcal{A}$ outputs two messages $m_0$ and $m_1$.
- A uniform bit $b$ is chosen, unknown to $\mathcal{A}$. $\mathcal{A}$ is given $c=E_k(m_b)$.
- $\mathcal{A}$ is not allowed to query the decryption oracle on $c$. $\mathcal{A}$ however continues to have oracle access to both encryption and decryption.
- $\mathcal{A}$ outputs bit $b'$. The result of the experiment is $1$, if $b=b'$, otherwise $0$.
So now, the scheme is considered CCA secure if the probability that the output of the above experiment is 1 is negligible. However, an adversary $\mathcal{A}$ could output $m_0=0^n, m_1=1^n$ and receive the challenge ciphertext $c=(r,s)$. It is not allowed to query $D_k(\cdot)$ on $c$. However, querying $D_k(c') = D_k \Big((r, s \oplus (0^{n-1}\|1)\;) \Big)$ would return $0^{n-1}\|1$ if $c=E_k(m_0)$, and $1^{n-1}\|0$ if $c=E_k(m_1)$.
Depending on the result of the query, the output of $\mathcal{A}$ is equal to $b$ with probability 1.