In today's knowledge based society, financial fraud has become a common phenomenon. Moreover, the growth in knowledge discovery in databases and fraud audit has made the detection of internal financial fraud a major area of research. On the other hand, auditors find it difficult to apply a majority of techniques in the fraud auditing process and to integrate their domain knowledge in this process. In this Paper a framework called "Knowledge-driven Internal Fraud Detection (KDIFD)" is proposed for detecting internal financial frauds. The framework suggests a process-based approach that considers both forensic auditor's tacit knowledge base and computer-based data analysis and mining techniques. The proposed framework can help auditor in discovering internal financial fraud more efficiently.