With　the　popularity　of　graph-structured　data　and　the　promulgation　of　various　data　privacy　protection　laws,　machine　unlearning　in　Graph　Convolutional　Network　(GCN)　has　attracted　more　and　more　attention.　However,　machine　unlearning　in　GCN　scenarios　faces　multiple　challenges.　For　example,　many　unlearning　algorithms　require　large　computational　resources　and　storage　space　or　cannot　be　applied　to　graph-structured　data,　and　so　on.　In　this　paper,　we　design　a　novel,　lightweight　unlearning　method　using　knowledge　distillation　to　solve　the　class　unlearning　problem　in　GCN　scenarios.　Unlike　other　methods　using　knowledge　distillation　to　unlearn　Euclidean　spatial　data,　we　use　a　single　retrained　deep　Graph　Convolutional　Network　via　Initial　residual　and　Identity　mapping　(GCNII)　model　as　the　teacher　network　and　the　shallow　GCN　model　as　a　student　network.　During　the　training　stage,　the　teacher’s　network　transfers　the　knowledge　of　the　retained　set　to　the　student　network,　enabling　the　student　network　to　forget　some　or　more　categories　of　information.　Compared　with　the　baseline　methods,　Graph　Unlearning　using　Knowledge　Distillation　(GUKD)　shows　state-of-the-art　model　performance　and　unlearning　quality　on　five　real　datasets.　Specifically,　our　method　outperforms　all　baseline　methods　by　33.77%　on　average　in　the　multi-class　experiments　on　the　Citeseer　dataset.　©　2023,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.