It　is　very　difficult　to　improve　Bayesian　classifier　detection　accuracy　because　selected　features　are　not　completely　independent.　In　this　paper,　unknown　malicious　code　detection　based　on　encoding　is　proposed　according　Minimum　Description　Length　(MDL)　principle.　We　use　encoding　module　instead　of　the　conditional　independence　assumption　and　statistical　classification　method.　The　experiment　results　show　our　method　is　efficient.