Social reputation (e.g., likes, comments, shares, etc.) on YouTube is the primary tenet to popularize channels/videos. However, the organic way to improve social reputation is tedious, which often provokes content creators to seek services of online blackmarkets for rapidly inflating content reputation. Such blackmarkets act underneath a thriving collusive ecosystem comprising core users and compromised accounts (together known as collusive users). Core users form the backbone of blackmarkets; thus, spotting and suspending them may help in destabilizing the entire collusive network. Although a few studies focused on collusive user detection on Twitter, Facebook, and YouTube, none of them differentiate between core users and compromised accounts. We are the first to present a rigorous analysis of core users in YouTube blackmarkets. To this end, we collect a new dataset of collusive YouTube users. We study the core-periphery structure of the underlying collusive commenting network (CCN). We examine the topology of CCN to explore the behavioral dynamics of core and compromised users. We then introduce KORSE, a novel graph-based method to automatically detect core users based only on the topological structure of CCN. KORSE performs a weighted k-core decomposition using our proposed metric, called Weighted Internal Core Collusive Index (WICCI). However, KORSE is infeasible to adopt in practice as it requires complete interactions among collusive users to construct CCN. We, therefore, propose NURSE, a deep fusion framework that only leverages user timelines (without considering the underlying CCN) to detect core blackmarket users. Experimental results show that NURSE is quite close to KORSE in detecting core users and outperforms nine baselines.