The　combination　of　deep　learning　and　intrusion　detection　has　become　a　hot　topic　in　today＇s　information　security.　In　today＇s　risky　network　environment,　the　ability　to　accurately　detect　anomalous　data　is　an　important　task　for　intrusion　detection.　In　an　intrusion　detection　system,　each　piece　of　data　contains　multiple　features.　However,　not　every　feature　will　determine　the　nature　of　the　data,　on　the　contrary,　too　many　features　will　affect　the　model＇s　judgment.　In　this　paper,　we　propose　an　intrusion　detection　model　of　a　deep　capsule　network　based　on　an　attention　mechanism.　The　model　uses　a　deep　capsule　network　to　enhance　the　extraction　of　features,　and　the　attention　mechanism　is　carried　out　to　make　the　model　focus　on　the　features　with　large　influences.　The　features　are　captured　in　multiple　directions　by　a　double　routing　algorithm　and　two　strategies　are　adopted　to　stabilize　the　dynamic　routing　process.　Finally,　experiments　are　conducted　on　the　intrusion　detection　dataset　with　good　results.