Efficient classification using parallel and scalable compressed model and its application on intrusion detection

摘要: Abstract In order to achieve high efficiency of classification in intrusion detection, a compressed model is proposed this paper which combines horizontal compression with vertical compression. OneR utilized as for attribute reduction, and affinity propagation employed select small representative exemplars from large training data. As be able computationally compress the larger volume data scalability, MapReduce based parallelization approach then implemented evaluated each step process abovementioned, on common but efficient methods can directly used. Experimental application study two publicly available datasets KDD99 CMDC2012, demonstrates that using effectively speed up detection procedure at 184 times, most importantly cost minimal accuracy difference less than 1% average.