Using an LDR, it will take two of them, the first one in series with the signal line and then a second one shunting the signal output from the attenuator, which would be the same as an "L-pad", functionally a voltage divider. The particular scheme for controlling the light intensity on each LDR will depend on the control intention required. No mention has been made of how the control command is initiated, nor has any mention of how many steps are desired. This might be an application with only two levels of attenuation.since you insist with the LDR I would like to know how you would do the circuit.
We do not have any real description of the desired operation aside from it being an attenuator. And why are only step reductions requested? One system that does allow remote control with only resistive devices would be a motor driven "T-Pad", which was included in one high-end audio device that I had worked with. Not cheap, but noise free and no power consumed except while changing positions.
And as for the non-linearity of the opto-isolation device described in the Farnell link, that is the transfer function of the package, it is not the current/voltage characteristic of the resistor portion of the device. LED operation is very non-linear, which is the way diodes work. But only the resistor would be in the audio portion of the circuit.