Adversarial attacks against text-independent speaker verification (TI-SV) systems assume access to genuine speaker's enrollment speech (e[n]). This assumption is self-defeating because if an attacker has e[n], they can bypass the system directly, making adversarial examples unnecessary. In contrast, we observe that the text-dependent SV (TD-SV) system, where the genuine speaker must say a password, offers a more practically relevant attack scenario. In reality, the attacker may not have access to the password spoken by a genuine speaker, but they can likely obtain normal speech from the genuine speaker. Therefore, generating adversarial noise that, when added to the genuine speaker's normal speech, can bypass the password requirement of a TD-SV system constitutes a potential realistic attack. This work investigates the feasibility of such a practical attack and shows that even the state-of-the-art TD-SV system is vulnerable with an attack success rate of 64.28 %.