Combining cryptography and steganography can provide more reliable and robust security by first concealing the existence of a secret message and then transforming it into ciphertext. While extensive research has been conducted on deep learning for steganalysis, no study to date has focused on a deep learning-based analysis of the combined use of cryptography and steganography. In this work, we introduce CS-Net, a deep learning-based approach to analyze the combined use of cryptography and steganography. To conceal the data, we used the well-known LSB steganographic algorithm. The concealed data is then encrypted using SPECK (-32/64), a lightweight encryption scheme developed by the NSA. We propose a novel approach that rotates the stego image containing concealed data to learn robust patterns in various directions of encryption. Further, we apply a state-of-the-art preprocessing technique used in steganalysis and demonstrate that it is still effective even when combined with cryptography. Indeed, even when the concealed data through steganography is encrypted (using SPECK), CN-Net distinguishes cover and stego images from the ciphertext. Our work generally achieves valid accuracy (greater than 0.5) for steganography/cryptography analysis.