Associative memory models are neural networks based on the mechanism of human memory. Memory content is expressed by activity of neural patterns, and the stored patterns are embedded in the synaptic couplings between the neurons. The stored information can only be retrieved when the initial state is enough near the stored information, but otherwise it can hardly be retrieved. In the (k, n) threshold scheme which is one of the secret sharing schemes, the secret information is divided into n shares, and then it is reconstructed only when k or more shares are assembled. We propose a secret sharing scheme constructed by the associative memory model, since there are some similarities between the associative memory model and the (k, n) threshold scheme. That is, the initial overlap of the associative memory model corresponds to the threshold of the secret sharing scheme. The secret information is stored in the associative memory model, and critical overlaps for number of division n and combination k are evaluated. As a result, we verified that the associative memory model worked as the (k, n) threshold scheme. Moreover, we found that some compress methods, i.e. Huffman method, would be useful for encoding the secret information to the stored pattern, since there is correlation between the stored patterns when text data are stored into the associative memory model.
展开▼
