Forensic analysis of images is used to detect various types of manipulations and forgeries. Median filters is one the anti-forensic methods which is used to smooth the visual characteristics of the altered image. Most of the forensic methods rely on detecting whether median-filtering or any other image restoration techniques are applied to the given image. However, image restoration techniques have a hard time making the statistical characteristics of the altered image similar to the original one. In this work, an anti-forensic method based on Generative Adversarial networks is presented, which is used to remove the traces of median filtering from the altered images and thus increases forensic undetectable. Through experiments, the proposed system achieved the improvement of 2.199 dB Peak Signal-to-Noise Ratio (PSNR) and 0.0279 Structural Similarity Index (SSIM) over the existing median filtering techniques.