This paper proposes a transform embedding method (STEM) which uses existing ideas from digital watermarking, improving upon them with empirical data to make them suitable for steganography.