The　Internet　of　Things　(IoT)　in　smart　cities　collects　and　transmits　a　large　amount　of　time-space-sensitive　information　to　realize　feedback　control.　It　bridges　the　gap　between　the　information　world　and　the　real　world.　With　the　data-based　feature,　the　security　and　credibility　of　the　IoT　mainly　depend　on　whether　the　source　of　the　data　is　trusted.　Therefore,　as　the　data　collection　and　transmission　entity,　sensing　nodes　should　be　classified　and　proved　the　trustworthiness.　However,　the　existing　works　failed　to　classify　and　measure　the　credibility　of　sensing　nodes　multidimensional　in　real　time.　The　previous　trust-proof　methods　also　cannot　effectively　protect　the　key　information.　To　address　these　problems,　this　article　first　proposes　a　multidimensional　and　fine-grained　dynamic　measurement　method　in　a　trusted　computing　environment.　Then,　a　trust　classification　model　of　sensing　nodes　is　presented,　and　a　grouping　mechanism　of　different　trust　levels　is　designed　to　identify　malicious　nodes.　Finally,　a　threshold　ring　signature-based　trust　certification　scheme　is　proposed　for　data　source　authentication.　It　can　adequately　protect　the　privacy　information　of　the　attestation　node　and　has　complete　anonymity　and　traceability.　Besides,　the　scheme　has　a　shorter　signature　and　high　computational　efficiency,　which　makes　it　also　suitable　for　sensing　nodes　with　limited　computing　resources.　The　simulation　results　show　that　the　scheme　has　better　dynamic　adaptability　and　can　effectively　ensure　the　credibility　of　data　sources　under　the　premise　of　various　attacks　with　accessible　impact　on　the　system.