Search

At recently, the most of the digital images are stored and transferred in their compressed format using discrete cosine transform (DCT)-based compression technique. DCT is one of the most important data compression technique due to the efforts from Joint Photographic Experts Group(JPEG). Blind steganalysis means how to detect the presence of the messages that are hidden using different types of steganography algorithms and has the ability to detect new unknown steganography algorithms. This paper presents blind steganalysis technique that can reliably detect the existence of messages hidden in JPEG files. In order to save the computation and memory cost, it is desirable to have image processing operations implemented directly in the DCT domain. The proposed method is based on extracting features directly from DCT domain through the analysis of differences between DCT coefficients before and after cropping. The extracted features are prepared as input to Support Vector Machine (SVM) to classify the image as stego (image that contain secret message) or clean (image that does not contain secret message). The experiments performed show that the proposed method yields better classification accuracy compared with other related works.

This paper presents a One-Time Pad (OTP) encryption and a genetic algorithm (GA) audio steganography, respectively. In the proposed scheme, genetic algorithm is used for both encryption and steganography. In the encryption process, the key is randomly generated using genetic algorithm. The generated key is used as a key for the One-Time Pad algorithm. Then, the encrypted message is embedded in audio in the best position for embedding using signal-to-noise ratio (SNR) in genetic algorithm fitness function. Results had been discussed for the proposed scheme. It dissolves the drawbacks of key generating and distribution for One-Time Pad by using genetic algorithm. The proposed method satisfies the main steganography requirements pointed in perceptual transparency, hidden data capacity and robustness. Experimental results had been made, namely text analysis, waveform analysis and signal-to-noise ratio. Besides, a comparison of SNR for the proposed method withsome known algorithm. All are to indicate that the proposed method can significantly outperform many steganography techniques in terms of security and audio quality.

We propose a novel neural network approach for the classification of abnormal mammographic images into benign or malignant based on their texture representations. The proposed framework has the capability of mapping high dimensional feature space into a lower-dimension, in a supervised way. The main contribution of the proposed classifier is to introduce a new neuron structure for map representation and adopt a supervised learning technique for feature classification. This is achieved by making the weight updating procedure dependent on the class reliability of the neuron. We showed high accuracy (95.2%) for our proposed approach in the classification of abnormal real mammographic images when compared to other related methods.

In this work, as part of our continuous effort, an optimized heuristic technique is proposed. The basic aim of this technique to fragment data vertically in Distributed Database System (DDBS). Drive by the queries along with using hierarchical clustering algorithm, the proposed technique seeks to propose an aggregated similarity measure to properly perform the clustering-based vertical data fragmentation. Data replication and allocation are also investigated to produce a comprehensive solution. As a matter of fact, the key concern is to find an effective best-fitting solution for improving DDBS throughput through developing an aggregated similarity based data fragmentation process, drawing a site clustering algorithm, and presenting a greedy-based transmission costs-reducing data allocation algorithm. Moreover, data replication is carefully considered in such a way that cost is to be essentially minimized.

In this paper, as integral part for our previous efforts, a heuristic model is proposed. The basic objective of this model is to partition data vertically in Distributed Database System (DDBS). Driven by the running queries along with using hierarchical clustering algorithm, the presented model pursues to find a method by which an efficient data fragmentation and allocation is set to be made. The prime focus of this paper is oriented towards shedding the light on previous studies as well as presenting greedy based transmission costs-reducing data allocation algorithm. Data replication is also investigated to produce a comprehensive solution. Moreover, the site clustering has been integrated into this work. Moreover, data replication is carefully considered in such a way that cost is to be essentially minimized.

present paper introduces destructive neural network learning techniques and presents the analysis of the convergence rate of the error in a neural network with and without threshold. Also, a constructive algorithm for rule extraction based on a trained neural network using Gene Expression Programming (GEP) is proposed. The rules are not an easy task due to the large number of examples entered to the input layer. Thus, we can use GEP to encode the rules in the form of logic expression. Finally, the proposed model is evaluated on different public-domain datasets and compared with standard learning models from WEKA, and then the results accentuate that the set of rules extraction from the proposed method is more accurate and brief compared with those achieved by the other models.

Imputation Techniques have been proposed mainly with the aim of predicting the missing values in the incomplete sets as an essential step in any data analysis framework. K-means-based Imputation, as a representative statistical imputation method, has been producing satisfied results in terms of effectiveness and efficiency in handling popular and freely available data set (e.g., Bupa, Breast Cancer, Pima, etc.). The main idea of K-means based methods is to impute the missing value relying on the prototypes of the representative class and the similarity of the data. However, such kinds of methods share the same limitations of the K- means as data mining technique. In this paper and motivated by such drawbacks, we introduce simple and efficient imputation methods based on K-means to deal with the missing data from various classes of data sets. Our proposed methods give higher accuracy than the one given by the standard K-means.

The artificial neural networks (ANNs), among different soft computing methodologies are widely used to meet the challenges thrown by the main objectives of data mining classification techniques, due to their robust, powerful, distributed, fault tolerant computing and capability to learn in a data-rich environment. ANNs has been used in several fields, showing high performance as classifiers. The problem of dealing with non numerical data is one major obstacle prevents using them with various data sets and several domains. Another problem is their complex structure and how hands to interprets. Self-Organizing Map (SOM) is type of neural systems that can be easily interpreted, but still can’t be used with non numerical data directly. This paper presents an enhanced SOM structure to cope with non numerical data. It used DNA sequences as the training dataset. Results show very good performance compared to other classifiers. For better evaluation both micro-array structure and their sequential representation as proteins were targeted as dataset accuracy is measured accordingly.

Steganography is one of the branches of information security field, it aims to hide information in unremarkable cover media so as not to arouse an eavesdropper’s suspicion. The secret message is hidden in such a way that no significant degradation can be detected in the quality of the original image. The aim of this paper is to introduce an efficient steganographic scheme to hide data over gray scale images. This scheme is based on the property of the human eye, which is more sensitive to the change in the smooth area than the edge area using pixel value difference, besides employing the LSB substitution technique as a fundamental stage. The experimental results show that the proposed method could successfully achieve the goals of the high embedding capacity and maintaining the visual quality, in addition, provides more secure data hiding using selective pixel positions determined by a secret image (i.e. key). Moreover, based on that, the secret message is replaced with dynamic LSBs, our scheme can effectively resist several image steganalysis techniques.

Imputation is a class of procedures that aims to fill the values which are missed with estimated ones. These methods involve replacing missing values with estimated ones based on some information available in the data set. K-means has been successful in finding missing values for several data sets available such as Bupa, Breast Cancer, Pima, etc. In this paper, we introduce an efficient imputation methods based K-means to treat missing data. Our proposed methods give higher accuracy than the one on given by classical K-means. Experimental results hold on a variety class of data sets.

Process mining refers to the extraction of process models from event logs. This paper presents a new process mining approach based on the combination of Honey Bee algorithm and Genetic Algorithm in which the benefits of Honey Bee algorithm is used where the concept of neighborhood search for a solution emerges from intelligent behavior of honeybee and the diversity of Genetic algorithm to find the global optimum. The new process mining approach presented in this paper is implemented as a plug-in in the process mining framework http://www.processmining.org. Computational experiments show that the process mining approach present in this paper gives a significant improvement over the basic Genetic algorithm.

Data security is one of major challenges in the recent literature. Cryptography is the most common phenomena used to secure data. One main aspect in cryptography is creating a hard to guess cipher. Artificial Neural Networks (ANN) is one of the machine learning techniques widely employed in several fields based on its characters, depending on the application area. One of these fields is data security. The state of art in this paper is the use of self organizing map (SOM) algorithm concept as a core idea to construct a pad; this pad is used to generate the cipher at one end. At the other end of communication the same process is synchronized to generate the same pad as the deciphering key. The security of the proposed model depends on the complex nature of ANN's. The algorithm could be categorized under symmetric cryptography, merging both stream and block cipher. A modified version of the same algorithm also presented employs permutation and variable SOM neighborhoods. The proposal can be applied over several file formats like videos, images, text files, data benchmarks, etc as show in experimental results.

In this paper, we propose two methods to improve the Ad-Hoc On-Demand Distance-Vector (AODV) protocol. The main goal in the design of the protocol was to reduce the routing overhead, buffer overflow, end-to-end delay and increase the performance. A multi-path routing protocol is proposed which is based on AODV and Ant Colony Optimization (ACO). This protocol is refereed to Multi-Route AODV Ant routing (MRAA). Also we propose a load balancing method that uses all discovered paths simultaneously for transmitting data. In this method, data packets are balanced over discovered paths and energy consumption is distributed across many nodes through network. This protocol is refereed to Load Balanced Multi-Route AODV Ant routing algorithm (LBMRAA).

In this paper, a new cryptosystem based on hybrid approaches is proposed. This scheme provides multi-security services. The proposed scheme takes the advantage of neural networks computation power to build an encryption algorithm to serve data confidentiality, where the encrypting process is done by used shard key as input to the ANN model, and the data need to be secured will be the output of it, the generated set of weights will act as the cipher. Also, made combination between MAC technique and RSA encryption algorithm produced digital signature of the scheme which provided the integrity and authentication services. The experimental results indicate that the proposed cryptosystem has a high confusion and diffusion properties, therefore it has a high security and it is suitable for secure communications.

In some medical applications where a tissue of interest covers a large fraction of the image or a prior knowledge on the region of interest is available, extracting features by fixed blocs in the image is sufficient. However in the general case, one would like to identify features for each tissue in the image. This would require prior image segmentation. Medical image segmentation is one of the most challenging problems in medical image analysis and a very active research topic. Therefore, there is no algorithm available in the general case for isolating medical image regions. This paper presents an accurate method for extracting texture features from medical image for classification. It is based on bloc wise clustering of medical images. The proposed technique extracts accurate and general set of texural features. Experimental result showed the high accuracy of the extracted textural features. Experiments held on mammographic image analysis society (MIAS) dataset.