A　fault　detection　filter　(FDF)　for　a　class　of　networked　systems　with　missing　measurements　is　studied.　The　packet　data　loss　is　modeled　as　a　Bernoulli　distributed　white　sequence　with　a　known　conditional　probability　distribution.　A　fault　detection　filter　is　then　designed　to　make　the　residual　error　system　exponentially　mean-square　stable　with　a　prescribed　disturbance　attenuation　level.　A　sufficient　condition　is　also　derived　in　terms　of　linear　matrix　inequality　(LMI).　Finally,　a　numerical　example　illustrates　the　effectiveness　of　the　proposed　approach.