Steganography is an art for invisible communication through message embedding in a host signal to evade detection by simple observation. Digital images, audio, video, etc. can act as host. There are many techniques for hiding message but one very common is embedding in least significant bit (LSB). Steganalysis deals with detecting the presence of data, if any, in a digital media. This paper proposes a reliable framework for detection of LSB Steganography in digital images. It is based on the principle that embedding data is equivalent to addition of noise in an image. Data embedding in a noise free image is likely to produce larger variations in the image characteristics in comparison to embedding in a noisy image. A close observation of these variations can help in determining if image is steganographic. In this paper, the difference between the close color pairs and unique color pairs is used as the distinguishing factor. Experimental results indicate that this method is capable of detecting presence of message embedding even when hidden message is small, i.e., ~2% of host size.