In　order　to　deliver　messages　between　moving　vehicles　in　cities,　this　paper　proposes　a　Social-based　Data　Forwarding　Mechanism　(SDFM)　for　V2V　communication　in　VANETs.　The　SDFM　learns　vehicles＇　social　characteristics　in　a　distributed　manner,　and　then　transfers　messages　in　a　＂Store-Carry-Forward＂　pattern.　SDFM　describes　the　sociality　of　a　vehicle　as　centrality　and　community.　Based　on　these　two　metrics,　for　each　message,　a　vehicle　forwards　it　to　the　vehicle　with　higher　global　centrality　than　itself　till　it　meets　a　vehicle　within　the　same　community　as　the　message＇s　final　destination,　and　then　forwards　the　message　to　the　vehicle　with　higher　local　centrality　till　it　is　finally　delivered.　The　SDFM　calculates　vehicles＇　centralities　by　computing　the　mobile　entropies　of　vehicles,　and　determines　vehicles＇　communities　by　adopting　a　distributed　K-Clique　algorithm　on　recorded　inter-meeting　time　as　well　as　meeting　frequencies　between　vehicles.　Experimental　results　show　that,　for　civic　moving　vehicles,　SDFM　mechanism　has　significantly　improved　the　message　delivery　ratio　and　the　average　delivery　delay　performances　when　compared　with　a　social　based　routing　protocol　bubble　rap　which　is　designed　for　data　forwarding　in　packet　switch　networks.