In order to solve the problems that IDSs and firewalls cannot efficiently detect new SQL injection and too much time is wasted when the security personnel reads log files to analyze attacks, we proposed and implemented a high-interaction web honey pot system for SQL injection analysis. By (i)modifying PHP extension for MySQL to intercept data-base requests and (ii)adopting exception based and signature based detection techniques, the system can generate the corresponding attack graphs to solve problems above. For illustration, SQL injection attack examples are utilized to show the performance of the honey pot system. The results show that the honey pot system can intercept all database requests and increase the efficiency of SQL injection analysis with the attack graphs. This system provides an efficient and timely detection on the new SQL injection and helps security personnel quickly analyzing the new SQL injection with the attack graph