A RANDOM PROJECTION APPROACH TO SECURE MEDICAL IMAGES

All our day?s work in this world of science is done with useful data. These useful data or information is extracted from raw facts. From the information, knowledge is gained. This knowledge is used by the customers for beneficial outcome. Here comes the concept of data mining. An important query arises as how to preserve these data. This concept is called as Privacy Preserving with Data Mining (PPDM). Many PPDM techniques are available to protect the data. The PPDM technique is useful in fields like medicine, forensics, defence, etc to preserve the confidential data. The existing techniques protect the secret data either by perturbing or by hiding them. Moreover, most of the techniques focus only on the numerical data. Very few perturbation techniques like translation, multiplicative and rotation perturb the images. But these techniques are very easily attacked by third parties since the transformation is a linear one. The security strength of these perturbation techniques is very low. Other perturbation techniques such as k- Anonymity, Rule Hiding and Data Swapping are applicable to numerical data. Excessive k-Anonymity also leads to data loss. Morphological operations greatly change the shape and structure of an image. But they are also reactive to noise and intrusions on the boundaries of the image. The privacy level of the images with these perturbation techniques is very low. The Research Work aims to overcome the drawbacks of the existing perturbation techniques. The objective of the Research Work is to improve the privacy level of the images and to strengthen the security of images by implementing a Random Projection technique. The scope of the Research Work is confined to medical images. Medical images with various dimensions such as 2D, 3D and 5D are taken. Dimensionality Reduction plays a key role in this Research Work. Dimensionality of the input images is reduced by applying Random Projection technique.