Asymmetric Sidel'nikov-type code-based cryptosystem is one of the generalizations of McEliece-type cryptosystems. The secret key in this generalization is the set of u randomly selected generator matrices GC 1 ,...,GC u of different [ni, k]-codes Ci, i ∈ {1, ..., u}. In addition, a part of the secret key is a random permutation ( u i=1 ni × u i=1 ni)-matrix P . The public key matrixG is the result of multiplying the concatenation of secret generator matrices and the matrix P :G = [GC 1 |...|GC u ]P . The security of these cryptosystems is based on the assumption that for u 2 it is computationally difficult to find inG such (k×ni)-submatrices composed of ni columns ofG, which would be generator matrices of codes permutably equivalent to the codes Ci, i = 1, ..., u. In the present paper, we construct an algorithm for the efficient search for such submatrices in the case when several conditions are satisfied. One of the conditions is the decomposability of the square of the connected codes C1,...,Cu into the direct sum of the squares of the codes Ci, i ∈ {1, ..., u}. An experimental assessment of the probability of fulfilling this condition for some Reed-Solomon codes, binary Reed-Muller codes, and Goppa code are also provided.