With　the　development　of　edge　computing　technology,　the　resource-limited　Internet　of　Things　(IoT)　devices　can　offload　computation-intensive　artificial　intelligence　tasks　such　as　model　training　and　inference　to　edge　servers　through　resource　trading.　However,　due　to　the　increase　in　the　number　of　intelligent　applications　and　the　rise　of　Peer-to-Peer　(P2P)　resource　trading,　the　existing　resource　trading　schemes　based　on　the　blockchain　can　no　longer　meet　the　needs　of　efficiency　and　security　at　the　same　time.　In　this　paper,　a　new　state　channels-based　resource　trading　scheme　is　proposed　for　IoT,　which　can　improve　scalability　without　sacrificing　security　and　fairness.　In　our　scheme,　most　of　the　trading　process　could　be　completed　off-chain,　and　the　blockchain　is　used　as　an　adjudicator　to　determine　malicious　behavior　according　to　the　users’　actions.　Moreover,　a　method　without　being　reliant　on　support　from　third　parties　is　presented　to　defend　against　execution　forks　that　must　be　faced　when　using　the　state　channels.　Finally,　the　feasibility　and　efficiency　of　our　scheme　are　experimentally　verified　in　the　realistic　testbed.　IEEE