In　order　to　detect　distributed　denial　of　service　(DDoS)　attack　effectively,　a　new　method　is　described.　Firstly,　Hilbert-Huang　transform　(HHT)　is　introduced　into　the　process　of　network　traffic　signal　to　obtain　corresponding　marginal　spectrum.　Then　its　Hurst　parameter　is　calculated.　Next,　the　detection　threshold　is　determined　by　means　of　training　Hurst　parameters　of　marginal　spectrums　of　experimental　data　with　the　goal　of　seeking　the　maximum　composite　index.　Finally　it　can　be　judged　whether　DDoS　attack　happened　or　not　through　comparison　with　above　threshold.　Experimental　results　verified　the　feasibility　of　this　method.