CHAPTER 8 CONCLUSION AND FUTURE SCOPE

Transcription

1 124 CHAPTER 8 CONCLUSION AND FUTURE SCOPE Data hiding is becoming one of the most rapidly advancing techniques the field of research especially with increase in technological advancements in internet and multimedia technology. With the technology advancing on one side, so has the rate of threat to hack, tamper or steal the data that is being transmitted over these medial also increased in leaps and bounds. This has necessitated an ever growing and systematic approach to ensure the security of transmitted and received content. On the other hand, the advent of tele based services have introduced medical image processing also to the data embedding arena especially for hospital management of patient records and subsequent follow up services and treatments. This maintenance of records also acts as data base for researchers all over to access any data as and when required. 8.1 SUMMARY OF WORK DONE In this thesis, a data hiding technique is proposed to embed electronic patient record (EPR) bits, multiple watermarks in the form of hospital logo and doctor s signature in a hybrid frequency domain technique addressing all the three key issues of any embedding process namely the robustness, visual imperceptibility and embedding capacity. The success of any data embedding technique lies on the above three factors often termed as optimality criteria or constraints for data hiding. They are rightly called constraints as they are interlinked with one another and improvement in one or more factor is done at the cost of degradation of the other. Since, our work focuses on medical images and its storage and retrieval at later stages, usage in real time for diagnosis and

2 125 subsequent treatments, a slight degradation in image quality could prove fatal for the patient. Hence, the goal in this work is to establish a balanced tradeoff between the three extremities for an optimal data embedding in the medical images. Chapters 3 and 4 focus on the first two optimality criteria namely robustness and visual imperceptibility. After careful and systematic analysis of all existing methodologies, a frequency domain strategy has been adopted due to its superiority in certain important aspects like robustness towards noise, resilience towards compression attacks, directionality etc., over the spatial domain techniques. A hybrid transform has been proposed after careful study of the features of the transforms in the hybrid combination so that it is able to address the robustness and visual imperceptibility criteria. Accordingly, the directional properties of the multi resolution Contourlet transform, the energy compaction property of DCT and the rotation - scaling translation invariance properties of the SVD transform have been joined together in a hybrid formation. The superiority of this combination is also tested and justified over other existing transforms like DCT, DWT in a single form or in a hybrid formation. The second aspect of this thesis is to address the third criterion which is embedding capacity. It is especially important since this application is directed towards hospital data management of patient records. Hence, it is desirable that maximum data measured in bits and bits per pixel are packed into the cover image without affecting the image quality. Existing methodologies have shown the variation of embedding capacity with image quality. However, it is a standalone approach as mentioned before. Since, the stored medical image and the patient records needs to be privately documented, stored and retrieved, a multiple watermarking technique has also been integrated into this thus ensuring security and also aiding in tamper detection and maintaining the legal authenticity of the record with respect to the patient, doctor and hospital. Hence, this proposed integrated data embedding system addresses the three optimality criteria which is a novel approach.

3 126 The final aspect of this thesis is to evaluate the strength, quality of the integrated data hiding system in terms of some important metrics such as peak signal to noise ratio, structural similarity index, and correlation coefficient and mean squared error. The evaluation is carried out by passing or exposing the embedded medical image to a wide range of aggressive image processing operations trying to simulate and equate to the real time attacks on the embedded image which may be intentional or unintentional. The third criterion is also evaluated by increasing the payload of the electronic patient information in stages and analyzing its signal quality in terms of peak signal to noise ratio. 8.2 CONTRIBUTION OF THE WORK Many researchers have put forward and proposed new and effective techniques of data embedding and extraction either by using a single transform or the transform in a hybrid combination. As mentioned in the previous section, up to the current knowledge, no technique has been found to address all the three issues in a single framework to a satisfactory balance. Hence, this work would be a useful and productive technique upon which further advancements and improvements could be done. The unique features of various transforms both in spatial and frequency domain have been made use of in development of this integrated data hiding system. Directional properties of Contourlet transform serve the purpose of choosing an ideal embedding location based on the energy level of the sub band, while the robustness properties of DCT have been utilized to provide the much needed resilience towards external attacks and the RST invariance of the SVD have been made use of to protect the underlying data against rotation and scaling attacks. On the other hand, the concept of histogram based difference expansion for increasing the embedding capacity acts as a useful tool especially for the proposed application of medical data management for patients and hospitals. Histogram shifting

4 127 jointly combined with difference expansion eliminates the need for a location map thus reducing the complexity. The ability of the images to be divided into smooth and non smooth regions facilitates the easy integration of robustness and visual imperceptibility property in the non smooth regions while accommodating the embedding capacity enhancement in the smooth regions of the image before concatenating them to get the embedded image which is quite robust, visually indistinguishable and with an increased payload. To summarize, flexibility exists in this technique to increase the payload at the expense of reduction in authentication procedures such as watermarks used for copyright protection and vice versa. A graphic user interface (GUI) would further enhance the simplicity of the method and further optimization if any needed could bring into picture genetic algorithm based techniques. 8.3 LIMITATIONS OF THE WORK Though a quite robust and high capacity data embedding system has been proposed for medical data management of patient records, data embedding is a fragile field as long as the level of threats and hacking techniques get complex day by day. As mentioned in previous sections, it can never reach saturation and needs to be updated with the latest hacking techniques and methodologies to keep the data embedding system strong and robust. In this work, it can be seen that the system loses much of its quality when cropped and scaled to higher limits. Data loss due to cropping could be minimized by appropriately choosing embedding locations which becomes contradictory to the high energy embedding location strategy adopted in this work. Moreover, data cropping is mostly intentional and hence it is difficult to predict the area of cropping before choosing the embedding location. In spite of the fact that a sufficient and satisfactory balance has

5 128 been brought about in the three optimality criteria, increase in payload beyond certain limits is found to impact image quality of cover image which is not quite desirable. 8.4 FUTURE SCOPE Data hiding has been an evergreen field since its inception in the late 1970 s with the ever increasing need for security, speed and low cost as part of requirements of any transmission or reception system. Things need to be done at a very rapid rate without any compensation in loss of data or time. Data hiding has proved to be the ideal choice for such applications. It has grown in leaps and bounds with the advent of broadband and communication technologies which is capable of spanning the entire globe within a few minutes. It has played key roles in covert communication, image and video tagging, broadcast monitoring and copyright protection. With the ever existing introduction of new transforms which are improvements over their predecessors in some way or the other also could contribute to the growth of data hiding in an attempt to further bring about the optimality. Directionality features of new transforms like the Curvelet transforms, Slantlet transforms could also be possibly investigated for its exploitation. As far as data hiding is concerned, there cannot be a point of saturation with the ever increasing threats and new hacking techniques in real world. With the advent of tele services, data hiding could cut costs and distance and travel miles and miles within a few minutes. It could also be an important tool in forensics for tracking criminal records and immediate identification of suspects. It could drastically reduce the effort of INTERPOL (International Police) and could bring about a better coordination globally. With the fast booming integrated circuit (IC) technology, data hiding concepts have been exposed to field programmable gate arrays (FPGA) using wavelet transforms, lifting based wavelet transforms. It could greatly aid in real time applications

6 129 with reduction in space, cost and time. This technology could be a great boon in preventing fake passports and forged identity cards especially in airline industries.