Calculating the reliability of PUF instance

[ahmad898]

Dear all, I have an Arbiter PUF with reliability of 99.58%. Suppose I have 256 input to the PUF which generates 256 responses as an output. Now, I want to replace the 50% of input bits with 50% of response bits, randomly. In fact, I create a feedback loop between the output and input of the Arbiter PUF. My question is that what is the reliability of my new PUF instance?
I believe that the reliability of the new PUF may be less than 99.58%, because the error may get propagated from the output to the input of PUF.

 

[dick_freebird]

I believe that your "reliability" number must incorporate some
specification of the incoming bit stream (particularly the relation
of data eye to clock position) while the outputs may be more
knowable, better aligned and thus not the same "unreliability
residue". If your outputs are stably well aligned then they ought
not to contribute further to the "residue"?

I imagine you'd want to take this apart for yourself, as it's only
conjecture at this end.