In the field of legacy software maintenance, there unexpectedly arise a large number of implicit coding ruleswhich are seldom written down in specification documents or design documentsas soflware becomes more complicated than it used be. Since not all the members in a maintenance team realize each of implicit ceding rules, a maintainer who is not aware of a rule often violates the rule while doing various maintenance activities such as adding new functionality and repairing faults. The problem here is not only such a violation causes injection of a new fault into software but also this violation will be repeated again and again in the future by different maintainers. Indeed, we found that 32.7% of t'au]ts of certain legacy software were due to such violations. This paper proposes a method for detecting code fragments that violate implicit coding rules. In the method, an expert maintainer firstly investigates the causes, situations, and code fragments of each fault described in bug reports; and, identifies implicit coding rules as much as possible. Then, code patterns violating the rules (which we call bug code patterns) are described in a pattern description language. Finally, potential faulty code fragments are extracted by a pattern matching technique.