Architecture-Driven Modernization (ADM) is a model-driven reengineering where systems are represented as instances of Knowledge Discovery Metamodel (KDM). KDM is the standard for representing systems in ADM context due to its power for capturing an extensive set of information about software systems. Besides, it is language and platform-independent, so every technique that is able of processing it also present this advantage. A recurrent activity in modernization projects is checking the conformance between the Current Architecture (CA) against the Planned Architecture (PA) in order to identify architectural drifts. The canonical phases of this activity are: (i) specification of the PA with its communication constraints; (ii) ex-traction of the CA, including the relationships among the architectural abstractions; and (iii) comparison between both architectures to identify the drifts. To the best of our knowledge, there is no ACC approach that addresses ACC in ADM context, considering KDM-represented systems. Therefore, we presents an ACC approach to be used in ADM context. We show how KDM can be used in ACC processes for representing the system to be modernized, the PA and the CA. We evaluated Arch-KDM using a small (LabSys-7KLoc) and a medium-size system (FreeMind-84KLoc) and the accuracy of the identification was acceptable.