Biblio

Deep learning is a highly effective machine learning technique for large-scale problems. The optimization of nonconvex functions in deep learning literature is typically restricted to the class of first-order algorithms. These methods rely on gradient information because of the computational complexity associated with the second derivative Hessian matrix inversion and the memory storage required in large scale data problems. The reward for using second derivative information is that the methods can result in improved convergence properties for problems typically found in a non-convex setting such as saddle points and local minima. In this paper we introduce TRMinATR - an algorithm based on the limited memory BFGS quasi-Newton method using trust region - as an alternative to gradient descent methods. TRMinATR bridges the disparity between first order methods and second order methods by continuing to use gradient information to calculate Hessian approximations. We provide empirical results on the classification task of the MNIST dataset and show robust convergence with preferred generalization characteristics.

This paper proposes an audio watermarking algorithm having good balance between perceptual transparency, robustness, and payload. The proposed algorithm is based on Cordic QR decomposition and multi-resolution decomposition meeting all the necessary audio watermarking design requirements. The use of Cordic QR decomposition provides good robustness and use of detailed coefficients of multi-resolution decomposition help to obtain good transparency at high payload. Also, the proposed algorithm does not require original signal or the embedded watermark for extraction. The binary data embedding capacity of the proposed algorithm is 960.4 bps and the highest SNR obtained is 35.1380 dB. The results obtained in this paper show that the proposed method has good perceptual transparency, high payload and robustness under various audio signal processing attacks.

A novel supervised machine learning system is developed to classify network traffic whether it is malicious or benign. To find the best model considering detection success rate, combination of supervised learning algorithm and feature selection method have been used. Through this study, it is found that Artificial Neural Network (ANN) based machine learning with wrapper feature selection outperform support vector machine (SVM) technique while classifying network traffic. To evaluate the performance, NSL-KDD dataset is used to classify network traffic using SVM and ANN supervised machine learning techniques. Comparative study shows that the proposed model is efficient than other existing models with respect to intrusion detection success rate.

Data's security being important aspect of the today's internet is gaining more importance day by day. With the increase in online data exchange, transactions and payments; secure payment and secure data transfers have become an area of concern. Cryptography makes the data transmission over the internet secure by various methods, algorithms. Cryptography helps in avoiding the unauthorized people accessing the data by authentication, confidentiality, integrity and non-repudiation. In order to securely transmit the data many cryptographic algorithms are present, but the algorithm to be used should be robust, efficient, cost effective, high performance and easily deployable. Choosing an algorithm which suits the customer's requirement is an utmost important task. The proposed work discusses different symmetric key cryptographic algorithms like DES, 3DES, AES and Blowfish by considering encryption time, decryption time, entropy, memory usage, throughput, avalanche effect and energy consumption by practical implementation using java. Practical implementation of algorithms has been highlighted in proposed work considering tradeoff performance in terms of cost of various parameters rather than mere theoretical concepts. Battery consumption and avalanche effect of algorithms has been discussed. It reveals that AES performs very well in overall performance analysis among considered algorithms.

In the literature, many chaotic systems have been used in the design of image encryption algorithms. In this study, an application of fractional order chaotic systems is investigated. The aim of the study is to improve the disadvantageous aspects of existing methods based on discrete and continuous time chaotic systems by utilizing the features of fractional order chaotic systems. The most important advantage of the study compared to the literature is that the proposed encryption algorithm is designed with a provable security approach. Analyses results have been shown that the proposed method can be used successfully in many information security applications.

In recent decades, a significant research effort has been devoted to the development of forensic tools for retrieving information and detecting possible tampering of multimedia documents. A number of counter-forensic tools have been developed as well in order to impede a correct analysis. Such tools are often very effective due to the vulnerability of multimedia forensics tools, which are not designed to work in an adversarial environment. In this scenario, developing forensic techniques capable of granting good performance even in the presence of an adversary aiming at impeding the forensic analysis, is becoming a necessity. This turns out to be a difficult task, given the weakness of the traces the forensic analysis usually relies on. The goal of this paper is to provide an overview of the advances made over the last decade in the field of adversarial multimedia forensics. We first consider the view points of the forensic analyst and the attacker independently, then we review some of the attempts made to simultaneously take into account both perspectives by resorting to game theory. Eventually, we discuss the hottest open problems and outline possible paths for future research.

This paper considers a pilot spoofing attack scenario in a massive MIMO system. A malicious user tries to disturb the channel estimation process by sending interference symbols to the base-station (BS) via the uplink. Another legitimate user counters by sending random symbols. The BS does not possess any partial channel state information (CSI) and distribution of symbols sent by malicious user a priori. For such scenario, this paper aims to separate the channel directions from the legitimate and malicious users to the BS, respectively. A blind channel separation algorithm based on estimating the characteristic function of the distribution of the signal space vector is proposed. Simulation results show that the proposed algorithm provides good channel separation performance in a typical massive MIMO system.

Live migration is the process used in virtualization environment of datacenters in order to take the benefit of zero downtime during system maintenance. But during migrating live virtual machines along with system files and storage data, network traffic gets increases across network bandwidth and delays in migration time. There is need to reduce the migration time in order to maintain the system performance by analyzing and optimizing the storage overheads which mainly creates due to unnecessary duplicated data transferred during live migration. So there is need of such storage device which will keep the duplicated data residing in both the source as well as target physical host i.e. NAS. The proposed hash map based algorithm maps all I/O operations in order to track the duplicated data by assigning hash value to both NAS and RAM data. Only the unique data then will be sent data to the target host without affecting service level agreement (SLA), without affecting VM migration time, application downtime, SLA violations, VM pre-migration and downtime post migration overheads during pre and post migration of virtual machines.

On account of large and inconsistent propagation delays during transmission in Underwater Wireless Sensor Networks (UWSNs), wormholes bring more destructive than many attacks to localization applications. As a localization algorithm, DV-hop is classic but without secure scheme. A secure localization algorithm for UWSNs- RDV-HOP is brought out, which is based on reputation values and the constraints of propagation distance in UWSNs. In RDV-HOP, the anchor nodes evaluate the reputation of paths to other anchor nodes and broadcast these reputation values to the network. Unknown nodes select credible anchors nodes with high reputation to locate. We analyze the influence of the location accuracy with some parameters in the simulation experiments. The results show that the proposed algorithm can reduce the location error under the wormhole attack.

Time and cost are the most critical performance metrics for computer systems including embedded system, especially for the battery-based embedded systems, such as PC, mainframe computer, and smart phone. Most of the previous work focuses on saving energy in a deterministic way by taking the average or worst scenario into account. However, such deterministic approaches usually are inappropriate in modeling energy consumption because of uncertainties in conditional instructions on processors and time-varying external environments. Through studying the relationship between energy consumption, execution time and completion probability of tasks on heterogeneous multi-core architectures this paper proposes an optimal energy efficiency and system performance model and the OTHAP (Optimizing Task Heterogeneous Assignment with Probability) algorithm to address the Processor and Voltage Assignment with Probability (PVAP) problem of data-dependent aperiodic tasks in real-time embedded systems, ensuring that all the tasks can be done under the time constraint with areal-time embedded systems guaranteed probability. We adopt a task DAG (Directed Acyclic Graph) to model the PVAP problem. We first use a processor scheduling algorithm to map the task DAG onto a set of voltage-variable processors, and then use our dynamic programming algorithm to assign a proper voltage to each task and The experimental results demonstrate our approach outperforms state-of-the-art algorithms in this field (maximum improvement of 24.6%).

Communication networks can be the targets of organized and distributed attacks such as flooding-type DDOS attack in which malicious users aim to cripple a network server or a network domain. For the attack to have a major effect on the network, malicious users must act in a coordinated and time correlated manner. For instance, the members of the flooding attack increase their message transmission rates rapidly but also synchronously. Even though detection and prevention of the flooding attacks are well studied at network and transport layers, the emergence and wide deployment of new systems such as VoIP (Voice over IP) have turned flooding attacks at the session layer into a new defense challenge. In this study a structured sparsity based group anomaly detection system is proposed that not only can detect synchronized attacks, but also identify the malicious groups from normal users by jointly estimating their members, structure, starting and end points. Although we mainly focus on security on SIP (Session Initiation Protocol) servers/proxies which are widely used for signaling in VoIP systems, the proposed scheme can be easily adapted for any type of communication network system at any layer.

Tensor decompositions, which are factorizations of multi-dimensional arrays, are becoming increasingly important in large-scale data analytics. A popular tensor decomposition algorithm is Canonical Decomposition/Parallel Factorization using alternating least squares fitting (CP-ALS). Tensors that model real-world applications are often very large and sparse, driving the need for high performance implementations of decomposition algorithms, such as CP-ALS, that can take advantage of many types of compute resources. In this work we present ReFacTo, a heterogeneous distributed tensor decomposition implementation based on DeFacTo, an existing distributed memory approach to CP-ALS. DFacTo reduces the critical routine of CP-ALS to a series of sparse matrix-vector multiplications (SpMVs). ReFacTo leverages GPUs within a cluster via MPI to perform these SpMVs and uses OpenMP threads to parallelize other routines. We evaluate the performance of ReFacTo when using NVIDIA's GPU-based cuSPARSE library and compare it to an alternative implementation that uses Intel's CPU-based Math Kernel Library (MKL) for the SpMV. Furthermore, we provide a discussion of the performance challenges of heterogeneous distributed tensor decompositions based on the results we observed. We find that on up to 32 nodes, the SpMV of ReFacTo when using MKL is up to 6.8× faster than ReFacTo when using cuSPARSE.

Image encryption takes been used by armies and governments to help top-secret communication. Nowadays, this one is frequently used for guarding info among various civilian systems. To perform secure image encryption by means of various chaotic maps, in such system a legal party may perhaps decrypt the image with the support of encryption key. This reversible chaotic encryption technique makes use of Arnold's cat map, in which pixel shuffling offers mystifying the image pixels based on the number of iterations decided by the authorized image owner. This is followed by other chaotic encryption techniques such as Logistic map and Tent map, which ensures secure image encryption. The simulation result shows the planned system achieves better NPCR, UACI, MSE and PSNR respectively.

The enormous size of video data of natural scene and objects is a practical threat to storage, transmission. The efficient handling of video data essentially requires compression for economic utilization of storage space, access time and the available network bandwidth of the public channel. In addition, the protection of important video is of utmost importance so as to save it from malicious intervention, attack or alteration by unauthorized users. Therefore, security and privacy has become an important issue. Since from past few years, number of researchers concentrate on how to develop efficient video encryption for secure video transmission, a large number of multimedia encryption schemes have been proposed in the literature like selective encryption, complete encryption and entropy coding based encryption. Among above three kinds of algorithms, they all remain some kind of shortcomings. In this paper, we have proposed a lightweight selective encryption algorithm for video conference which is based on efficient XOR operation and symmetric hierarchical encryption, successfully overcoming the weakness of complete encryption while offering a better security. The proposed algorithm guarantees security, fastness and error tolerance without increasing the video size.

This article presents a systematic review on the challenges and recent progress of timing and carrier synchronization techniques for high-speed optical transmission systems using single-carrier-based coherent optical modulation formats.
 

One Time Password which is fixed length strings to perform authentication in electronic media is used as a one-time. In this paper, One Time Password production methods which based on hash functions were investigated. Keccak digest algorithm was used for the production of One Time Password. This algorithm has been selected as the latest standards for hash algorithm in October 2012 by National Instute of Standards and Technology. This algorithm is preferred because it is faster and safer than the others. One Time Password production methods based on hash functions is called Hashing-Based Message Authentication Code structure. In these structures, the key value is using with the hash function to generate the Hashing-Based Message Authentication Code value. Produced One Time Password value is based on the This value. In this application, the length of the value One Time Password was the eight characters to be useful in practice.
 

A robust adaptive filtering algorithm based on the convex combination of two adaptive filters under the maximum correntropy criterion (MCC) is proposed. Compared with conventional minimum mean square error (MSE) criterion-based adaptive filtering algorithm, the MCC-based algorithm shows a better robustness against impulsive interference. However, its major drawback is the conflicting requirements between convergence speed and steady-state mean square error. In this letter, we use the convex combination method to overcome the tradeoff problem. Instead of minimizing the squared error to update the mixing parameter in conventional convex combination scheme, the method of maximizing the correntropy is introduced to make the proposed algorithm more robust against impulsive interference. Additionally, we report a novel weight transfer method to further improve the tracking performance. The good performance in terms of convergence rate and steady-state mean square error is demonstrated in plant identification scenarios that include impulsive interference and abrupt changes.

The gradient-descent total least-squares (GD-TLS) algorithm is a stochastic-gradient adaptive filtering algorithm that compensates for error in both input and output data. We study the local convergence of the GD-TLS algoritlun and find bounds for its step-size that ensure its stability. We also analyze the steady-state performance of the GD-TLS algorithm and calculate its steady-state mean-square deviation. Our steady-state analysis is inspired by the energy-conservation-based approach to the performance analysis of adaptive filters. The results predicted by the analysis show good agreement with the simulation experiments.

Dynamic firewalls with stateful inspection have added a lot of security features over the stateless traditional static filters. Dynamic firewalls need to be adaptive. In this paper, we have designed a framework for dynamic firewalls based on probabilistic ontology using Multi Entity Bayesian Networks (MEBN) logic. MEBN extends ordinary Bayesian networks to allow representation of graphical models with repeated substructures and can express a probability distribution over models of any consistent first order theory. The motivation of our proposed work is about preventing novel attacks (i.e. those attacks for which no signatures have been generated yet). The proposed framework is in two important parts: first part is the data flow architecture which extracts important connection based features with the prime goal of an explicit rule inclusion into the rule base of the firewall; second part is the knowledge flow architecture which uses semantic threat graph as well as reasoning under uncertainty to fulfill the required objective of providing futuristic threat prevention technique in dynamic firewalls.

In this work, a new fingerprinting-based localization algorithm is proposed for an underwater medium by utilizing ultra-wideband (UWB) signals. In many conventional underwater systems, localization is accomplished by utilizing acoustic waves. On the other hand, electromagnetic waves haven't been employed for underwater localization due to the high attenuation of the signal in water. However, it is possible to use UWB signals for short-range underwater localization. In this work, the feasibility of performing localization for an underwater medium is illustrated by utilizing a fingerprinting-based localization approach. By employing the concept of compressive sampling, we propose a sparsity-based localization method for which we define a system model exploiting the spatial sparsity.

The main focus of this work is the estimation of a complex valued signal assumed to have a sparse representation in an uncountable dictionary of signals. The dictionary elements are parameterized by a real-valued vector and the available observations are corrupted with an additive noise. By applying a linearization technique, the original model is recast as a constrained sparse perturbed model. The problem of the computation of the involved multiple parameters is addressed from a nonconvex optimization viewpoint. A cost function is defined including an arbitrary Lipschitz differentiable data fidelity term accounting for the noise statistics, and an ℓ0-like penalty. A proximal algorithm is then employed to solve the resulting nonconvex and nonsmooth minimization problem. Experimental results illustrate the good practical performance of the proposed approach when applied to 2D spectrum analysis.