مجله رایانش نرم و فناوری اطلاعات
سال یکم شماره 2 (تابستان 1391)

Gravitational Search Algorithm (GSA) is a novel searching and optimization algorithm which has been reported recently. GSA was inspired by the gravitational forces between the mechanical objects. The movements of the searching objects in this method are based on the estimated accelerations and velocities of them. In this paper utilizing of GSA is investigated for unsupervised image clustering. At first an improvement for the conventional GSA is presented and then an appropriate fitness function is defined for unsupervised image clustering. Extensive experimental results on various data and images demonstrate the high performance of the proposed method in comparison to other algorithms.

Fuzzy classification systems play an important role in dealing with uncertainty and vagueness inherent in multi-dimensional pattern classification problems. Finding an optimal fuzzy rule set is a milestone in designing fuzzy classification systems. In this paper, a fuzzy Genetic Algorithm (GA) is developed to generate fuzzy classification rules. The proposed algorithm produce an optimal rule set in terms of number of rules and accuracy. Hence, in this research the optimal rule set consists of appropriate number of fuzzy rules in order to maximize the number of correctly classified patterns. In order to illustrate the efficiency of the proposed method, it is applied to Iris dataset and a collection of data contained in 2D images captured by a video camera to identify the class of moving objects in a traffic scene. The simulation results are also compared with other methods.

Analysis of social networks is a useful tool for providing information about social behaviors. One of the objectives achieved in analyzing social networks is to detect malicious behaviors and protect people privacy that results in providing security in social networks. There are different approaches for anomaly detection in social networks that can be categorized into two groups, parametric and non-parametric methods. In this paper, a new hybrid approach is presented which uses both parametric and nonparametric approaches. Features like continuous data variability and the inference which results in variety of conditions can be analyzed based on fuzzy logic. Furthermore the complexity of social networks needs a structured method that can simplify the analysis process, so cliques are evaluated as base structures for anticipating malicious intent in social networks. The proposed approach is used in empirical case study to detect malicious users and a comparison is made with previous approaches.

In this paper, for secure transmission of speech signal in wireless sensor, a new method based on wavelet transform has been proposed. In this methodology, speech signal is compressed and encrypted simultaneously using characteristics of wavelet transform. The output speech which is in form of an obscure bit stream is required to be inserted in carrier signal. For insertion, fivestage wavelet transform is applied on speech signal and the optimum sub-bands are selected. The simulation results show that the proposed method provides higher security, lower complexity and lower consuming energy in comparison with popular methods in wireless sensor networks.

Orthogonal Frequency Division Multiplexing (OFDM) is a multicarrier modulation technique which is used for supporting high data rates. In this modulation scheme, Fourier transform is used to provide orthogonality between subcarriers. Since the channel is multi-path, a cyclic prefix having a length greater than channel delay spread is added to each OFDM symbol. The cyclic prefix removes the inter symbol interference, but, since it does not carry useful information, it decreases spectrum efficiency. In wavelet-based OFDM, there is no need for cyclic prefix. Wavelets have suitable localization, finite length and orthogonality both in time and frequency domains. Therefore, they could be used to generate suitable waveform for transmission through a fading channel. In this paper, a comparison has been made between performances of the Fourier transform and orthogonal wavelet families in OFDM-based systems in order to improve. Also, in order to select optimum filter order in these systems, the effect of increasing filter order is investigated. Simulation results show that wavelet transforms outperform the Fourier transform in terms of bit error rate (BER) and system efficiency and wavelet transforms with smaller filter orders should be used.

Security is one of the most important issues in modern computer systems. Among the major challenges in these systems is identifying normal and abnormal behaviors. But the boundary between these two is not well defined and it is a very complicated task to accomplish. Intrusion detection systems are one of the techniques that are used to maintain security in computer networks. Incorrect report of the intrusion alarm system is one of the major problems of security systems. Intrusion is defined as a set of activities and the purpose of these activities is jeopardizing the integrity, reliability and unauthorized system access to a particular resource.This paper proposes a mechanism based on the decision tree technique to detect intrusions in the network. Simulation results show that by the proposed method not only the duration of the training phase was reduced but also the detection rate and false alarm rate were improved.