Gigabit Passive Optical Networks (GPONs) provide high bandwidth and scalability as well as their architecture features enable the telecommunication operators to cover larger areas with fewer active elements, which reduces cost. However, to provide excellent customer experience and quality of service, the detection of faults in GPONs needs to be fast and efficient, allowing administrators to respond to network problems quickly. Moreover, the wide variety of equipment and the amount of data exchanged by them makes the fault diagnosis a quite challenging task. To overcome these issues, this paper proposes agNOsTic sOftware aRchItecture fOr aUtomated diagnoSis (NOTORIOUS), an architecture model that makes it possible to discover faults in GPONs employing a rule-based correlation with different scopes. Hence, our evaluations show use of the proposed solution can increase network availability, allowing the customer to experience a better quality of service.