You might want to read Cryonics, cryptography, and maximum likelihood estimation.

Short summary: if cryptanalytic methods can recover the wiring of World War II rotor machines knowing only some input-output pairs and with only limited information about the actual wiring, then similar algorithms should be able to recover the neuronal "wiring" between different cortical areas when we already have a wealth of information about that wiring plus a good knowledge of acceptable input-output pairs.