The paper considers a methodology for encrypting color images, in which the key is the initial state of an elementary cellular automata for the implementation of evolution based on a given rule. The research task is to improve the reliability of encryption of color images based on the encryption of the bit layers that make up the image. To solve this problem, the methodology of forming the evolution of an elementary cellular automata was used, which is a finite bit key array for each bit layer of the image. The encryption and decryption key consists of subkeys, the number of which corresponds to the number of bits that encode the color of each pixel. Each subkey consists of the initial states of an elementary cellular automata and the rules that shape its evolution. For the formation of each bit key array, different initial conditions and different Wolfram rules for elementary cellular automata were used. The size of each formed key bit array is equal to the size of the corresponding bit layer of the color image. Encryption is performed by using the XOR function for the generated key bit array and bit-slice of the image. As a result of the experiments, it was established that it is necessary to use different rules that form different geometric shapes in evolution. It is also established that it is necessary to form a key bit array for each bit layer starting not from the first lines of evolution. It was found that the quality of encryption of a color image is most influenced by the three most significant bits of each byte of the pixel code, encoding the corresponding red, blue and green colors.