اسم المؤلف:
فرقد حامد عبد الرحیم
اسم المشرف:
ستار بدر سدخان المالكي
المستخلص:
Security evaluation algorithms can be considered as one of the most important challenges in computer networks. This is because of the growing data sharing among all clients (users). Therefore, the security level evaluation aspect of cryptography systems is recently appeared to be very important.In this work, evaluations of (RSA and AES) encryption methods are carried out by using Fuzzy Inference System (FIS) and Adaptive Neuro - Fuzzy Inference System (ANFIS). The editor of MATLAB (2013) is employed in this study and it contains a hybrid ANFIS facility between the Artificial Neural Network (ANN) and Fuzzy Logic techniques.First of all, designing and programming software codes for the first encryption method (RSA) have been simulated according to its original algorithm. Consequently, executing the RSA algorithm to collect the data values is implemented for the following parameters (message length, execution time, length of key and cipher message entropy). These parameters have been considered in the proposed approaches. So, the RSA data is used as the bases of the FIS inputs. Then, all the training and testing data values have been collected from the proposed FIS and prepared to be used in the next step (the ANFIS). The number of training samples has been selected to be 100 values by executing special software programs. These values have been utilized as follows : opening the ANFIS editor; loading the training data; determining the main ANFIS parameters and training the data with the least error tolerance. Subsequently, the number of testing samples has been chosen to be also 100 values by implementing special software programs. Hence, the evaluations are observed and the characteristics of the ANFIS which attained the best tested results have been benchmarked. Similar steps to evaluate the RSA by using large key numbers are implemented except of utilizing the parameter (key length) to study the influence of the key value on security evaluations. The proposed FISSecurity evaluation algorithms can be considered as one of the most important challenges in computer networks. This is because of the growing data sharing among all clients (users). Therefore, the security level evaluation aspect of cryptography systems is recently appeared to be very important.In this work, evaluations of (RSA and AES) encryption methods are carried out by using Fuzzy Inference System (FIS) and Adaptive Neuro - Fuzzy Inference System (ANFIS). The editor of MATLAB (2013) is employed in this study and it contains a hybrid ANFIS facility between the Artificial Neural Network (ANN) and Fuzzy Logic techniques.First of all, designing and programming software codes for the first encryption method (RSA) have been simulated according to its original algorithm. Consequently, executing the RSA algorithm to collect the data values is implemented for the following parameters (message length, execution time, length of key and cipher message entropy). These parameters have been considered in the proposed approaches. So, the RSA data is used as the bases of the FIS inputs. Then, all the training and testing data values have been collected from the proposed FIS and prepared to be used in the next step (the ANFIS). The number of training samples has been selected to be 100 values by executing special software programs. These values have been utilized as follows : opening the ANFIS editor; loading the training data; determining the main ANFIS parameters and training the data with the least error tolerance. Subsequently, the number of testing samples has been chosen to be also 100 values by implementing special software programs. Hence, the evaluations are observed and the characteristics of the ANFIS which attained the best tested results have been benchmarked. Similar steps to evaluate the RSA by using large key numbers are implemented except of utilizing the parameter (key length) to study the influence of the key value on security evaluations. The proposed FISSecurity evaluation algorithms can be considered as one of the most important challenges in computer networks. This is because of the growing data sharing among all clients (users). Therefore, the security level evaluation aspect of cryptography systems is recently appeared to be very important.In this work, evaluations of (RSA and AES) encryption methods are carried out by using Fuzzy Inference System (FIS) and Adaptive Neuro - Fuzzy Inference System (ANFIS). The editor of MATLAB (2013) is employed in this study and it contains a hybrid ANFIS facility between the Artificial Neural Network (ANN) and Fuzzy Logic techniques.First of all, designing and programming software codes for the first encryption method (RSA) have been simulated according to its original algorithm. Consequently, executing the RSA algorithm to collect the data values is implemented for the following parameters (message length, execution time, length of key and cipher message entropy). These parameters have been considered in the proposed approaches. So, the RSA data is used as the bases of the FIS inputs. Then, all the training and testing data values have been collected from the proposed FIS and prepared to be used in the next step (the ANFIS). The number of training samples has been selected to be 100 values by executing special software programs. These values have been utilized as follows : opening the ANFIS editor; loading the training data; determining the main ANFIS parameters and training the data with the least error tolerance. Subsequently, the number of testing samples has been chosen to be also 100 values by implementing special software programs. Hence, the evaluations are observed and the characteristics of the ANFIS which attained the best tested results have been benchmarked. Similar steps to evaluate the RSA by using large key numbers are implemented except of utilizing the parameter (key length) to study the influence of the key value on security evaluations. The proposed FISapproach confirmed that the RSA evaluation is successfully implemented to the ANFIS editor.All the previous steps are repeated for the AES encryption method except one difference. That is, the utilized parameters here are the (message length, execution time and cipher message entropy). Basically, two key values are determined for the AES, which equals to 128 bits. Likewise the RSA, the suggested procedures are applied to the AES and the proposed FIS approach confirmed that the AES evaluation is successfully implemented to the ANFIS editor.Finally, comparisons between this study and previous work, and between the RSA and AES are established. In addition, comparisons between the evaluated outcomes of the FIS and ANFIS have been investigated by using two statistical metrics.
اسم المؤلف:
حيدر عباس محسن
اسم المشرف:
حازم باقر طاهر العلي
المستخلص:
Fractal Image Compression (FIC) is one of the lossy techniques. In order for fractal compression image to be encoded is partitioned into non - overlapping blocks called ranges. From this range pool creates a new array called domain pool, the data of domain is array produced from taking an average of every four (2x2) adjacent elements in the range array, the domain is divided into overlapping blocks. Each block in range pool should be matched with every large number of blocks in the domain pool in order to find minimum error and recorded its best IFS approximate. The decoder process applies the determined IFS transformations on any initial image, and the process is repeated many times until reaching the attractor. In this thesis, it presents reducing the long encoding time of FIC without making significant sacrifice in the quality. In addition to the traditional fractal compression method, four IFS coding methods have been tested.The first method is speed up encoding phase by using moment features of the domain blocks with threshold , this method it compute the moment features of domain blocks only , during the matching process, compares the ratio factor of domain block with threshold, then compute the scaling (s), offset (o) coefficients and minimum error between range and domain blocks, and storing the optimal IFS coding , the results of this method with lenna image (4x4) block size, (the encoding time = 50.29 sec), (compression ration= 4.57 ) ,(PSNR=24.04), (RTR=33.42 ). The second method is speed up encoding phase by using the entropy technique of the domain blocks only , during the matching process, compares entropy value of domain block with threshold, and storing optimal IFS coding , the results of lenna image(4x4) block size, (the encoding time = 95.47sec), (compression ration= 4.47 ) ,(PSNR= 24.08) and (RTR= 17.60) . In the third method it merges the entropy technique with moment features of domain blocks only , during the matching process, compares entropy value and moments ratio factor of domain blocks with threshold and storing IFC coding, the compression results of lenna image (4x4)block size , (encoding time = 100.90 sec), (compression ration=4.47) ,(PSNR= 24.11) and (RTR=16.66) , the flaw in this method is the encoding time of 16x16 partitioning more than 4x4 and 8x8 partitioning. In the fourth method speed up encoding phase by use moment features of the range and domain blocks, in this method it computes the moment features of range and domain blocks, during the matching process, if the moments ratio factor of range and domain blocks are equal then compute scale(s),offset(o) and error (R) .storing best IFS coding. the results of lenna image(4x4) (encoding time = 114.02 sec), (compression ration= 4.47) ,(PSNR= 24.43), (RTR=14.74), the flaw in this method is the encoding time of 16x16 partitioning more than 4x4 and 8x8 partitioning. Matlab R2008a was used as a program to designed and implemented to achieve the coding and testing tasks. The environment used for our tests is a single PC with processor (core i3 CPU 2.53GHz) and 4.00 GB RAM .
