In one implementation, the disclosure provides systems and methods for a multi-party secret sharing protocol that is device specific in that the secret matrix used herein is tied to individual computing devices. Specifically, the method includes determining device channel errors of a plurality of computing devices based on channel impulse response (CIR) of communication channels of the plurality of computing devices, training a linear regression model using the device channel errors to generate learning with error (LWE) secrets for each of the plurality of computing devices, generating a general access structure secret matrix using the LWE secrets from each of the plurality of computing devices, and distributing shares of the general access structure secret matrix to the plurality of computing devices based on a multi-party secret sharing protocol, wherein the multi-party secret sharing protocol provides that the general access structure secret matrix cannot be constructed without shares from an authorized set of the computing devices.