An improved algorithm with key attributes constraints for mining interesting association rules in network log

Computer logs are generated by application activities, network accesses and system audit, which are important data sources for user pattern mining, computer forensic analysis, intrusion detection analysis and outlier detection. Algorithms for mining association rule are useful methods to find interesting rules implied in large computer log data. But existing algorithms which based on confidence and support are unfit for mining computer log data, many uninteresting rules will be generated and useful rules will be shadowed. To solve this problem, the concept of key attributes of network log data is introduced, and an algorithm with key attributes constraints for mining interesting association rules in network log data is designed. Experimental result shows that the number of uninteresting rules can be reduced effectively and the validity of rules which mined are improved.