As　the　core　of　moving　target　defense,　defense　strategy　has　been　the　focus　of　research.　However,　the　existing　defense　strategies　of　moving　targets　defense　lack　pertinence　and　still　belong　to　the　real-time　and　passive　defense　state.　In　order　to　realize　the　advance　deployment　of　defense　measures,　network　traffic　should　be　accurately　predicted　in　the　long-term.　In　this　paper,　a　traffic　prediction　model　based　on　segment　lifting　wavelet　transform　and　LSTM　network　is　proposed　(LSTM-SLWT).　Furthermore,　the　distance　between　the　reconstructed　traffic　and　the　predicted　approximate　sequence　is　calculated　by　using　the　dynamic　time　warping　algorithm,　and　the　distance　is　considered　as　the　possible　DDoS　attack　intensity.　Then　the　genetic　algorithm　is　used　to　select　the　mutation　element　variation　range　that　can　carry　out　effective　defense.　So　as　to　verify　the　effectiveness　of　the　proposed　method,　this　paper　conducted　a　prediction　experiment　on　the　traffic　of　different　time　lengths.　Experimental　results　show　that　the　prediction　effect　of　LSTM-SLWT　is　better　than　that　of　LSTM　and　the　traditional　LSTM-LWT　prediction　model.　Moreover,　the　effective　defense　strategy　is　selected　on　the　basis　of　traffic　prediction,　which　can　reduce　the　defense　cost　while　ensuring　the　security　of　the　system.　©　2022　IEEE.