In this paper we propose a new JPEG-based steganography technique, using eigenvalues’ properties of matrices of quantized DCT coefficients. We call this new approach, Eigenvalue Steganography. Usual steganography techniques based on JPEG images, hide data by changing the least significant bit (LSB) of DCT coefficients. Eigenvalue Steganography technique hides confidential data employing the eigenvalues of quantized DCT matrices and manipulating them. This technique suggests the subdivision of quantized DCT coefficients blocks into smaller submatrices in DCT domain. Afterwards, this technique efficiently hides confidential information among DCT submatrices, by employing the proposed algorithms and manipulating the eigenvalues of these submatrices. Experiments done in this paper, depicts that the two proposed methods to implement Eigenvalue Steganography show appropriate resistance against Subtractive Pixel Adjacency Matrix (SPAM) steganalyzer. This steganalysis model doesn’t have the capability of proper detection for the suggested methods.