Search

Dementia is a neurocognitive brain disease that emerged as a worldwide health challenge. Machine learning and deep learning have been effectively applied for the detection of dementia using magnetic resonance imaging. In this work, the performance of both machine learning and deep learning frameworks along with artificial neural networks are assessed for detecting dementia and normal subjects using MRI images. The first-order and second-order hand-crafted features are used as input for machine learning and artificial neural networks. And automatic feature extraction is used in the last framework with the pre-trained networks. The outcomes show that the framework using the deep neural networks performs better contrasted with the first two methodologies used in terms of various performance measures.

Background: Software Product Line is the group of multiple software systems which share the similar set of features with multiple variants. Feature model is used to capture and organize features used in different multiple organization. Objective: The objective of this research article is to obtain an optimized subset of features which are capable of providing high performance. Methods: In order to achieve the desired objective, two methods have been proposed. a) An improved objective function which is used to compute the contribution of each feature with weight based methodology. b) A hybrid model is employed to optimize the Software Product Line problem. Results: Feature sets varying in size from 100 to 1000 have been used to compute the performance of the Software Product Line. Conclusion: The results shows that proposed hybrid model outperforms the state of art metaheuristic algorithms.

Objectives: The modern science applications have non-continuous and multivariate nature due to which the traditional optimization methods suffer a lack of efficiency. Flower pollination is a natural interesting procedure in the real world. The novel optimization algorithms can be designed by employing the evolutionary capability of the flower pollination to optimize resources. Methods: This paper introduces the hybrid algorithm named Hybrid Hyper-Heuristic Flower Pollination Algorithm, HHFPA. It uses a combination of Flower Pollination Algorithm (FPA) and Hyper- Heuristic Evolutionary Algorithm (HypEA). This paper compares the basic FPA with the proposed algorithm named HHFPA. FPA is inspired by the pollination process of flowers whereas the hyper-heuristic evolutionary algorithm operates on the heuristics search space that contains all the heuristics to find a solution for a given problem. The proposed algorithm is implemented to solve the Quality of Service (QoS) based Service Composition Problem (SCoP) in Internet of Things (IoT). With increasing services with same functionality on the web, selecting a suitable candidate service based on non-functional characteristics such as QoS has become an inspiration for optimization. Results: This paper includes experimental results showing better outcomes to find the best solution using the proposed algorithm as compared to Basic FPA. Conclusion: The empirical analysis also reveals that HHFPA outperformed basic FPA in solving the SCoP with more convergence rates.

Software Product Line (SPL) customizes software by combining various existing features of the software with multiple variants. The main challenge is selecting valid features considering the constraints of the feature model. To solve this challenge, a hybrid approach is proposed to optimize the feature selection problem in software product lines. The Hybrid approach ‘Hyper-PSOBBO’ is a combination of Particle Swarm Optimization (PSO), Biogeography-Based Optimization (BBO) and hyper-heuristic algorithms. The proposed algorithm has been compared with Bird Swarm Algorithm (BSA), PSO, BBO, Firefly, Genetic Algorithm (GA) and Hyper-heuristic. All these algorithms are performed in a set of 10 feature models that vary from a small set of 100 to a high-quality data set of 5000. The detailed empirical analysis in terms of performance has been carried out on these feature models. The results of the study indicate that the performance of the proposed method is higher to other state-of-the-art algorithms.

The Internet of Things (IoT) is a novel technology that has opened doors to the new level of interaction between the things. This has resulted in an enhancement in the quality of life and optimized use of various resources. IoT uses various technologies related to networking, sensing, databases and artificial intelligence to enhance the lifestyle and makes business processes simpler. In the IoT based systems, the number of devices contributing to a particular application is very large, spreading to a large geographical area at various locations. The device can be considered as a resource used by a service in the application layer. IoT service is the most important entity. The real-world things generate data by perceiving the environment in large quantity and store it in the distributed databases or cloud databases. This manuscript analytically and statistically categorizes and analyze the current research techniques on the service discovery and selection in the IoT, published between 2010 and 2018. It finds that the discovery among a huge number of services requires fast, scalable and dynamic service discovery mechanism over the Internet. Once the services are discovered, the next step is to select the most appropriate service. This paper includes a comprehensive analysis of the discovery and selection of services in IoT. A patent related to service and discovery have been also discussed which would be beneficial in identifying the research gap to make the system more efficient.

Web service compositions are commendable in structuring innovative applications for different Internet-based business solutions. The existing services can be reused by the other applications via the web. Due to the availability of services that can serve similar functionality, suitable Service Composition (SC) is required. There is a set of candidates for each service in SC from which a suitable candidate service is picked based on certain criteria. Quality of service (QoS) is one of the criteria to select the appropriate service. A standout amongst the most important functionality presented by services in the Internet of Things (IoT) based system is the dynamic composability. In this paper, two of the metaheuristic algorithms namely Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) are utilized to tackle QoS based service composition issues. QoS has turned into a critical issue in the management of web services because of the immense number of services that furnish similar functionality yet with various characteristics. Quality of service in service composition comprises of different non-functional factors, for example, service cost, execution time, availability, throughput, and reliability. Choosing appropriate SC for IoT based applications in order to optimize the QoS parameters with the fulfillment of user’s necessities has turned into a critical issue that is addressed in this paper. To obtain results via simulation, the PSO algorithm is used to solve the SC problem in IoT. This is further assessed and contrasted with GA. Experimental results demonstrate that GA can enhance the proficiency of solutions for SC problem in IoT. It can also help in identifying the optimal solution and also shows preferable outcomes over PSO.

In this paper, a model is presented for classification of Dementia brain disease using magnetic resonance imaging. Bag of features (BOF) is used for extracting the features of MRI scans, which are classified using multi-class Support Vector Machine (SVM) for distinguishing the scans into three categories as demented, mildly cognitive impaired and normal controls. The Magnetic Resonance Images (MRI) used for the classification is obtained from ADNI database. The speeded up robust features are extracted using BOF approach. The experimental results showed the achievement of competitive performance in terms of classification accuracy using the proposed methodology. The BOF approach clubbed with SVM leads to an accuracy of 93%.

Internet of Things (IoT) provides an innovative worldwide view to the things which are fit for communicating with one another. Each thing like temperature sensor, pressure sensor provides its functionality in the form of web service on the Internet. As the number of things connecting to the IoT network are increasing, thus one of the important challenges is to address the optimal selection of the web service to fulfil the user request. As the user requirements cannot be fulfilled using a single service so various services are composed to serve the IoT based system applications. This paper presents the usage of multi-objective metaheuristic search algorithm to provide optimal solution to Service Composition Problem (SCP) in IoT. Furthermore, this paper illustrates the use of one of the popular algorithms namely Non-dominated Sorting Genetic Algorithm (NSGA II). This paper looks at the way and methodology utilized to solve SCP in IoT. QoS parameters have become an important criterion for optimization. This algorithm is utilized to solve Quality of Service (QoS) based SCP. The results obtained are evaluated by analysing the no. of points plotted by pareto optimal curve. NSGA II is used to obtain the pareto fronts.

Background: Intrusion detection systems are responsible for detecting anomalies and network attacks. Building of an effective IDS depends upon the readily available dataset. This dataset is used to train and test intelligent IDS. In this research, NSL KDD dataset (an improvement over original KDD Cup 1999 dataset) is used as KDD’99 contains huge amount of redundant records, which makes it difficult to process the data accurately. Methods: The classification techniques applied on this dataset to analyze the data are decision trees like J48, Random Forest and Random Trees. Results: On comparison of these three classification algorithms, Random Forest was proved to produce the best results and therefore, Random Forest classification method was used to further analyze the data. The results are analyzed and depicted in this paper with the help of feature/attribute selection by applying all the possible combinations. Conclusion: There are total of eight significant attributes selected after applying various attribute selection methods on NSL KDD dataset.

Proper quality and supervision of fruits during harvesting would be beneficial for the buyer and also result in yielding better income for the supplier. Agricultural products are graded on the basis of their dimension and quality. Traditionally, grading was human dependent; however, grading of fruits by human is not efficient, work-intensive and prone to errors. We propose an automated grading system for sorting tomatoes based on their maturity level. Not only does this speed up the processing time, but it also minimizes the error. A great demand for this is observed both in foreign and local markets. Our research utilizes openly available datasets for fruit grading, uses neural network-based advanced computer vision and transfer learning techniques and achieves a state-of-the-art accuracy in classifying maturity for all of the tomato images from the test set correctly. The framework suggested in this study could aid in precisely grading tomatoes based on ripeness and help agricultural industries in various processes like automatic segmentation of tomatoes on conveyor belts, or robotic harvesting, thereby projecting to impact hundreds of millions of tonnes of food that get wasted each year due to maturity factors at harvest time.

The paper focuses on bringing improvization in existing summarization algorithms to get deeper insights into properties of text that gives an edge to explore new relations of sentences in text and helps attain high accuracy. The goal is to amalgamate the existing network graph theory and its properties with the existing algorithms to evolve and resolve the unanswered questions and discover new insights to improve the precision and exactness. The thought is to develop a novel approach that discovers new ways to summarize using network summarization using significant similarity measures that indeed plays a significant role. The paper presents a network embedding approach that helps to find nearest neighbors in the embedding space and make recommendations or clusters depending on similarity. The network graph theory provides an edge for visualization of concepts and discovers the relations between different categories.

About 1 in 11 people around the world have some kind of diabetes, which is one of the leading causes of death. The two major types of diabetes which are prominent are type 1 and type 2 diabetes, where about 90% of the total cases are related to type 2. Fortunately, people with diabetes can lead long and healthy lives when their diabetes is diagnosed early, well managed, and self-monitored. Self-monitoring data can help people manage their diabetes and prevent hyperglycemic (high glucose levels) and hypoglycemic (low glucose levels) conditions. It is important to note that glucose-level fluctuations vary from individual to individual based on their current condition and other biological factors. In this research, we propose a methodology to manage diabetes using data from continuous glucose monitoring (CGM) devices and predicting glucose levels after a certain activity like intake of meals, intensive workouts, sleep, etc. takes place. The methodology comprises (1) analysis of ARIMAX and FbProphet time series forecasting techniques to predict future glycemic levels on the Nightscout dataset and (2) sheds some light over the interpretability of other parameters such as carbohydrates and insulin. The experimental results show that the FbProphet model outperforms the ARIMAX model.

Dementia is a brain disorder which causes loss of memory leading to disruption in the normal course of life of an individual. It is emerging as a global health problem in adults of age 65 years or above. The early diagnosis of dementia has gone forth as a key research zone having an aim of early identification for hindering the advancement. Artificial Intelligence, Machine Learning and Deep Learning have acquired a great importance in the medical field. Deep Learning provides path breaking application in medical imaging. This study gives a detailed summary of different implementation of deep learning for detecting the disease Dementia.

Life Expectancy is of great interest for any country so that proper economic and social status can be analyzed and the population groups in maximum need gets the benefit. This paper investigates the effect of the various representative parameters which impact life expectancy. Around 193 countries data are analyzed through visualization techniques to bring out the relationship between different parameters which have impact on life expectancy. Relationship between dependent variable (life expectancy) and various other independent variable is established through five regression techniques. The hyperparameter tuning for these techniques has been done through Grid search method. The Gradient Boosting Regression gives the best results and is also capable of overcoming overfitting.

Nowadays, there has been recent interest in applying machine learning techniques to the neurodegenerative disorders. Dementia is one such emerging global health issue and its early detection can be very helpful. A comparative analysis of four machine learning algorithms i.e., J48, Naive Bayes, Random Forest and Multilayer Perceptron is presented in this paper. For the attribute reduction, the CFSSubsetEval is used. It has been shown that J48 is performing best among all the algorithms for the detection of Dementia.

Diabetes, a chronic disease affects various organs of human body including the retina. Diabetic Retinopathy (DR) results from the Diabetes Mellitus (DM). In literature various machine learning algorithms have been applied in detection of DR. This involves two steps; Feature extraction and Classification. This paper reviews the various techniques used for detecting DR based on the features like blood vessels, microaneurysms, haemorrhages etc. In most of the experiments retinal fundus images were used in which images of retina were captured by fundus camera. This review bifurcates the detection of DR into two approaches; Blood vessels segmentation and Identification of lesions. This paper compares the experimental results of various machine learning techniques based on parameters like sensitivity, specificity, Area Under Curve (AUC), Accuracy. The results are also compared with the deep neural networks and analysis of best technique has been provided.

Universal Steganalysis rely on extracting higher order statistical features that gets disturbed when hiding the message in a clean image. Due to content adaptive steganographies like HUGO, WOW etc. which embed the data more in textured areas of the image rather than smooth areas by minimizing the distortion of the image itself, first order features are not sufficient to differentiate clean and stego images. Thus, rich models come into picture in which a large number of features are extracted based on higher order noise residuals of clean and stego images. Thus, Universal Steganalyser is essentially a supervised classifier built on high dimensional feature set. To work with such high dimensional features on a large dataset of images is a very challenging task due to curse of dimensionality as well as computationally very expensive. This paper aims at comparing performance of three techniques-Ensemble classifier, Logistic regression and K-Nearest Neighbors on Spatial Rich Model features extracted for benchmarked dataset BOSSbase_1.01, for the better discrimination of clean and stego images.

Typically, the Internet of Things (IoT) incorporates the provisioning of powerfully adaptable and virtualized things such as cameras, sensors as valuable resources over the Internet. There are different types of services that can be combined to meet user demands. Capturing the user requirements and composing various services from the existing set of services is required in order to facilitate the adaption of an IoT based application. The industry has a strong demand for service composition mechanisms in order to meet the fast and growing demands of the user on the web. There are various techniques and methodologies which can be applied to provide the optimized solution to the problem of Service Composition(SC) in IoT environments. This study involved a survey of existing methodologies that can be addressed to make effective SC in IoT.

This paper presents an improved firefly algorithm (DyFA) for feature selection that improves the convergence rate and reduces computational complexity through dynamic adaptation in blind image steganalysis. The alpha and gamma parameters of the Firefly algorithm are made to vary dynamically with each generation for faster convergence. If firefly algorithm’s performance does not improve for certain numbers of iterations then the particles with the worst fitness function values are replaced with new particles in the search space and particle dimensions are reduced by eliminating redundant features. This approach is effective in reducing computational complexity and improving detection capability of the classifier. To further reduce the computational complexity a hybrid DyFA is designed by ensemble of a filter approach (t test + regression) and wrapper approach (DyFA) incrementally. In this study, support vector machine classifier with radial basis function kernel and ten fold cross validation is used to evaluate the effectiveness of the proposed Firefly algorithm. DyFA is compared with well-known wrapper feature selection algorithms. Experimental results are performed on datasets constructed from four steganography algorithms nsF5, Perturbed Quantization, Outguess and Steghide with subtractive pixel adjacency matrix (SPAM) feature vector from spatial domain and Cartesian Calibrated features extracted by Pevnýfeature vector from transform domain. Experimental results demonstrate that DyFA reduces computation time and improves classification accuracy as compared to other feature selection algorithms. Hybrid DyFA shows an improvement in classification accuracy and in eliminating redundant features in more than 85 % of cases with respect to hybrid GLBPSO.

Dementia is a brain disorder emerging as a global health problem in adults of age 65 years or above. A lot of studies are going on for the early detection of dementia using various machine learning algorithms. In this study, we have implemented four machine learning algorithms as classifiers namely, Naive Bayes, Random Forest, Multilayer Perceptron and SMO in combination with various feature selection techniques to obtain the classification accuracy as a measure for the detection of dementia. It is found that Random Forest is showing best result of 98.6% when CFSSubsetEval Feature selection technique is used.

The key task of Steganalyzer is to identify if a carrier is carrying hidden information or not. Blind Steganalysis can be tackled as two-class pattern recognition problem. In this paper, we have extracted two sets of feature vectors from discrete wavelet transformation domain of images to improve performance of a Steganalyzer. The features extracted are histogram features with three bins 5, 10, and 15 and Markov features with five threshold values 2, 3, 4, 5, 6, respectively. The performance of two feature sets is compared among themselves and with existing Farid discrete wavelet transformation features based on parameter classification accuracy using neural network back-propagation classifier. In this paper, we are using three Steganography algorithms outguess, nsF5 and PQ with various embedding capacities.

This paper proposes a novel feature selection approach to improve the classification accuracy and reduce the computational complexity in image steganalysis. It is a hybrid filter-wrapper approach based on improved Particle Swarm Optimization (PSO). It consists of two phases: the first phase is composed of two filter techniques namely t test and multiple-regression which selects the features based on their ability to discriminate images as stego or cover. The second phase further reduces the number of features by working on the significant features selected during the first phase using an improved PSO. This approach overcomes the disadvantages of global best PSO by integrating it with local best PSO and dynamically changing the population size (Hope/Rehope). The proposed approach is tested on two sets of features extracted from spatial domain (SPAM-Subtractive Adjacency Matrix) and transform domain (CCPEV-Cartesian Calibrated features extracted by Pevný) for four embedding algorithms nsF5, Outguess, Perturbed Quantization and Steghide using SVM (Support Vector Machine) classifier. Experimental results demonstrate that this approach significantly improves the classification accuracy and drastically reduces dimensionality as compared to results produced by some well-known feature selection algorithms.

this paper, we are comparing 72 DWT feature set with 274 DCT feature set on the basis of parameters MSE (Mean square error), Accuracy and time for Blind Image Steganalysis. Dataset is created consisting of cover images and stego images, which are obtained by using two steganography tools steghide and outguess.Extracted features are then fed to Neural network Back propagation classifier, to compare the performance.Experimental results show that DCT feature set has higher performance as compared to DWT feature set.

This paper highlights the importance of wrapper feature selection technique for image steganalysis. Two major processes in blind image steganalysis are feature extraction and classification. To increase the efficiency of classifier feature space has been increased causing “curse of dimensionality”. This leads to requirement of feature selection for steganalysis. This paper discusses the importance of heuristic wrapper approach which is an improved version of Particle Swarm Optimization. The fitness function and classification is based on artificial neural network. The neural networks are fast learning and non-linear classifier which gives it an edge over other classifiers. The experiment results show that feature selection technique improves the classification accuracy with reduced number of feature set and provides better understanding of features.

The role of software product line (SPL) is very important in representing the same system with multiple variants. Feature models are used to define SPL. In this paper, genetic algorithm (GA), hyper-heuristic algorithm and particle swarm optimisation (PSO) have been applied for feature selection optimisation in SPL. Also, an improved fitness function is applied for optimisation of features in SPL. The objective function is designed by taking reusability and consistency of features (components) into consideration. Furthermore, we have used a case study and discussed about software product line in detail. A non-parametric test, i.e., Kruskal-Wallis test has been performed to analyse performance and computation time of 20 to 1,000 features sets and identify core features. Through extensive experimental analysis, it is observed that PSO outperforms GA and hyper-heuristic algorithm.

Objectives: Intrusion detection is the need of technical world where data is generating and changing at a very rapid rate. In last decade feature selection is the science that has given a new perspective to research in the area of Intrusion Detection System. Objective of this paper is to perform an analysis and comparison of various feature selection techniques with a new technique of hybrid Particle Swarm Optimization (PSO). Statistical Analysis: In this paper well known filter and wrapper feature selection techniques have been explored along with a hybrid PSO technique on the standard KDDCup99 dataset. A comparative analysis is performed over four filter techniques and two wrapper based techniques. Four different classifiers are compared to select the one providing good accuracy on the dataset. Findings: The hybrid PSO feature selection technique gives significant improvement in prediction capability as compared to traditional feature selection approaches. Analysis shows that SVM classifier provides better classification results. SVM is used as classifier because of its high accuracy. The analysis over 4 filter and two wrapper techniques shows that Hybrid PSO provides better results with 98.6% accuracy and 24 feature subset. Application/Improvements: Analysis provides importance of hybrid PSO, which may be applied to not only intrusion detection but also various other areas where feature reduction is required.

Objectives: To evaluate the performance of different tools that acquire, analyze and recover the evidences of crime from volatile memory. A comparison between different tools is presented with the aim of generating better understanding of the tools employed. Methods: Volatile memory stays for a very short period and that is why it is always hard to analyze such memory. It contains much useful information such as passwords, usernames, running processes, etc. Acquiring, analyzing and recovering are the three major steps for memory forensics. Experiments are performed with different tools to understand the procedure of acquiring, analyzing and recovering important evidences. Findings: The strengths and drawbacks of all the tools are analyzed that providesa better understanding of the working of the tools in specific scenarios. The tools like FtkImager and Belkasoft represent the data as a tree structure which makes it difficult to analyze the data. All the tools investigated are not entirely fitted for a particular situation hence; the investigation needs to rely on many tools that can retrieve useful information from the evidences. It is important to know the usefulness of a tool before it is applied to solve a crime. Although most of the tools are successful in providing reasonable evidence, no single tool is sufficient to complete the investigation. Improvements: Most of the tools work as passive agents that is it is left to the discretion of the investigator to analyze the evidences collected through different tools. The tools can be improved by combining it with machine learning techniques. This paper also discusses the improvements that can be done in order to make the working of the tools easier and yielding better results.

Steganalysis is capable of identifying the carrier(s) which have information hidden in them in such a way that their very existence is concealed. In this paper we propose a classification system with neural networks which reduces computational complexity through a pre-processing step (feature selection) performed by Bhattacharyya distance for image steganalysis. This approach is able to identify relevant features which are a subset of original features extracted from spatial as well as transform domain. It helps in overcoming the problem of “curse of dimensionalty” by removing redundant features by feature selection step before classifying the dataset. The experiments are performed on dataset obtained by four steganography algorithms outguess, steghide, PQ and nsF5 with two classifiers Support Vector Machine and Back Propagation neural networks. Classifier in combination with Bhattacharyya distance filter feature selection approach shows an improvement of 2–20% against total number of features.

Steganalysis aims at detecting the presence of steganography with or without knowledge of steganographic technique used. Feature based Steganalysis determines the presence of secret data by comparing the statistical features of both cover and stego images. Using a large number of features for Steganalysis relative to training set size may reduce classification accuracy and also increase computational complexity. Feature Selection selects subset of features from the given dataset which maximizes classifiers performance. In this paper we have applied four different filter feature selection approaches to features extracted from transform domain and then compared their performance by using five different classifiers.

This paper introduces a novel steganography-based approach to protect the iris data by hiding it into a digital image for personal identification purpose. Transformations are done to enrypt the biometric data before hiding. JPEG quantization blocks in diagonal sequence are investigated for exploring an efficient hiding algorithm with high transparency and strong robustness. The extraction of iris data does not require original image and provide high accuracy under different attacks. Experimental results show that this proposed method improves the security of the iris-feature with hardly detectable decrease in recognition performance.

The objective of forensic steganalysis is to identify the existence of embedded message and to ultimately retrieve the secret message. In this paper, we have extracted Histogram, Markov and Co-occurrence features from wavelet domain and compared with existing farid 72 DWT features. The performance metrics are MSE (Mean square error) and classification accuracy for blind image steganalysis using two steganography algorithms outguess and nsF5. Neural network back propagation classifier has been used to classify images into stego images and clean images. Experimental results show that the proposed features considerably outperform existing farid 72 DWT features. Keywords — Discrete wavelet transformation, Neural network, Outguess, Steganography, Steganalysis, nsF5. I. INTRODUCTION Steganography[1] is a process in which the message is embedded in a cover medium to create a stego medium with the help of steganography tool that results in changes in the statistical properties of cover medium where medium could be video, text, image, audio etc. Now a day’s wide variety of steganographic tools are freely and widely available on the internet such as StegHide [2], Outguess [3] etc. Statistical un-detectability [4] is the necessity of steganography process, which means that it is difficult or impossible for attacker to judge whether an image is the stego or cover based on the statistics. Steganography hides data in [5] Spatial domain and Transform domain or Frequency domain. Transform Domain is further divided into three categories (a) DCT (Discrete Cosine Transformation technique), (b) DWT (Discrete Wavelet transformation technique), (c) DFT (Discrete Fourier transformation technique).In transform domain, images are initially transformed and then the message is embedded while in spatial domain message is embedded directly inside the pixel.

Feature Selection is a preprocessing technique with great significance in data mining applications that aims at reducing computational complexity and increase predictive capability of a learning system. This paper presents a new hybrid feature selection algorithm based on Discrete Firefly optimization technique with dynamic alpha and gamma parameters and t-test filter technique to improve detectability of hidden message for Blind Image Steganalysis. The experiments are conducted on important dataset of feature vectors extracted from frequency domain, Discrete Cosine Transformation and Discrete Wavelet Transformation domain of cover and stego images. The results from popular JPEG steganography algorithms nsF5, Outguess, PQ and JP Hide and Seek show that proposed method is able to identify sensitive features and reduce the feature set by 67% in DCT domain and 37% in DWT domain. The experiment analysis shows that these algorithms are most sensitive to Markov features from DCT domain and variance statistical moment from DWT domain.

Steganography is the art of hiding the secret messages in an innocent medium like images, audio, video, text, etc. such that the existence of any secret message is not revealed. There are various Steganography tools available. In this paper, we are considering three algorithms - nsF5, PQ,Outguess. To compare the robustness and to withstand the steganalytic attack of the above three algorithms, an algorithm based on sensitive features is presented. SVM and Neural Network Pattern Recognition Tool is used on sensitive features extracted from DCT domain. A comparison between the accuracy obtained from SVM and NPR is also shown. Experimental results show that the Outguess method can withstand steganalytic attack by a margin of 35% accuracy as compared to nsF5 and PQ, hence Outguess is more reliable for Steganography.