With　the　rapid　development　of　autonomous　driving　techniques,　the　cyber-attack　issue　has　emerged　as　a　serious　concern　due　to　the　cyber　vulnerability　of　autonomous　vehicles.　In　this　paper,　we　consider　the　cyber　security　for　the　autonomous　vehicular　flow　which　suffers　from　the　cyber-attack　signals　on　speed　dynamics.　To　address　this　problem,　the　Aw-Rascle-Zhang　model　is　extended　by　considering　the　unknown　cyber-attack　to　the　autonomous　vehicles.　A　novel　simultaneous　state　and　input　estimation　algorithm　is　developed　for　the　cyber-attack　ARZ　model　by　using　the　boundary　observer.　The　exponential　stability　of　the　simultaneous　estimation　algorithm　is　given　with　a　set　of　inequality　conditions.　Finally,　numerical　simulation　illustrates　that　the　proposed　algorithm　can　estimate　the　traffic　flow　state　and　unknown　cyber-attack　simultaneously.