Sensitive information requires safe and secure transmission. To hide such information, cryptographic models are used which require a set of keys in their working. Maintaining such keys itself is costly and securing them adds to the task of security. Also, such models are specific to a particular type of information that can be hidden i.e. not dynamic in working. The proposed interface omits the use of these complex mathematical models instead uses Neural Networks for data hiding. In this technique during the training phase of the network, it is subjected to a variety of data and once trained can be utilized to encrypt or decrypt the information of any size irrespective of what was used in training set. This all being done at senders end, it is now transmitted over any channel to the designated receiver. For better results and analysis we incorporated our focus on the type of network, number of neurons, number of layers, simplicity and on parameters such as Peak Signal to Noise Ratio (PSNR). The proposed method shows greater flexibility in operation and low maintenance in regard to other crypto methods. The suggested technique would be suitable to incorporate in scenarios requiring least suspicion during an exchange of information and hence less prone to attackers. It would be easy to utilize on public modes of communication as sensitive information would be hidden in any media file which will appear just like some other normal media file.