This　study　proposes　a　multiple　linear　regression　architecture　based　on　stream　homomorphic　encryption　computing　to　analyze　ciphertext　for　massive　secure　data　computing.　The　proposed　architecture　contains　three　subsystems　including　terminal　subsystem,　data　access　subsystem,　and　data　computing　subsystem.　The　method　used　behind　the　presented　architecture　contains　four　stages　which　are　data　preprocessing　stage,　data　access　stage,　data　computing　stage,　and　result　processing　stage.　In　the　practical　experiments,　a　case　study　of　traffic　information　prediction　was　selected　to　evaluate　the　proposed　system　and　method.　The　predicted　traffic　information　was　generated　by　the　proposed　method　in　accordance　with　the　encrypted　traffic　information.　Our　experimental　results　showed　that　the　proposed　architecture　can　effectively　and　promptly　obtain　the　predicted　traffic　information.