This　paper　proposes　an　incomplete　information　Bayesian　Stackelberg　game　which　is　adapted　to　the　Cognitive　Internet　of　Vehicles　(CIoV)　network　to　defend　against　Spectrum　Sensing　Data　Falsification　(SSDF)　attacks　from　Malicious　Vehicle　Users　(MVUs).　Specifically,　this　paper　considers　the　random　appearance　of　MVUs　caused　by　mobility,　intelligent　SSDF　attacks　of　MVUs,　and　the　different　spectrum　sensing　performance　among　Vehicle　Users　(VUs).　In　the　game,　the　Fusion　Center　(FC)　as　the　leader　aims　to　improve　the　global　detection　performance　while　effectively　identifying　the　identities　of　different　VUs　by　optimizing　the　global　decision　threshold　and　the　reputation　threshold.　On　the　other　hand,　this　paper　models　the　random　appearance　of　MVUs　as　a　Poisson　random　process,　and　the　MVUs　are　the　intelligent　followers,　they　optimize　the　attack　probabilities　according　to　the　FC’s　strategies　to　evade　detection,　increase　the　chance　of　selfish　transmission　and　the　damage　to　the　CIoV　network.　To　solve　the　MVUs’　non-convex　optimization　problem,　this　paper　uses　the　Successive　Convex　Approximation　(SCA)　technique　to　obtain　MVUs’　optimal　attack　probabilities.　For　the　FC,　this　paper　proposes　the　method　combining　alternating　optimization　and　SCA　to　solve　the　non-convex　optimization　problem　of　the　FC　and　obtain　its　optimal　defense　strategies.　This　paper　also　proves　the　convergence　of　the　proposed　method　and　existence　of　the　Stackelberg　Equilibrium　(SE).　The　simulation　results　demonstrate　the　validity　and　superiority　of　the　proposed　method　compared　with　traditional　methods.　IEEE