Vehicular　Ad　hoc　Networks　(VANETs)　play　an　increasingly　important　role　in　a　number　of　applications,　particularly　those　associated　with　location-based　services　(e.g.,　spatial　keyword　searches　-　SKS).　However,　there　is　a　need　to　strike　a　balance　between　privacy　guarantee　and　search　efficiency,　and　existing　SKS　solutions　cannot　be　directly　implemented　in　VANETs.　In　addition,　existing　schemes　may　also　lack　support　for　semantic-awareness　in　the　dynamic　setting.　To　address　these　limitations,　we　propose　a　Secure　Semantic-aware　Spatial　Keyword　Search　scheme　that　supports　Dynamic　update　(3SKSD).　Specifically,　we　leverage　the　Latent　Dirichlet　Allocation　(LDA)　topic　model　and　secure　k　Nearest　Neighbor　(k)　method　to　help　us　achieve　both　efficiency　and　security.　We　also　construct　an　encrypted　R-tree　structure　to　facilitate　SKS　and　dynamic　update.　Moreover,　we　propose　an　advanced　scheme　with　forward　security　on　the　basis　of　3SKSD,　with　the　aim　of　minimizing　privacy　leakage　due　to　dynamic　updates.　Our　formal　security　analysis　verifies　the　validity　and　security　of　3SKSD,　and　findings　from　the　experimental　evaluation　demonstrate　its　high　search　accuracy　and　efficiency.