Authors: Vyshnav, M.T. | Sachin Kumar, S. | Mohan, Neethu | Soman, K.P.
Article Type: Research Article
Abstract: The present paper proposes Random Kitchen Sink based music/speech classification. The temporal and spectral features such as spectral centroid, Spectral roll-off, spectral flux, Mel-frequency cepstral coefficients, entropy, and Zero-crossing rate are extracted from the signals. In order to show the competence of the proposed approach, experimental evaluations and comparisons are performed. Even though both speech and music signals differ in their production mechanisms, those share many common characteristics such as a common spectrum of frequency and are comparatively non-stationary which makes the classification difficult. The proposed approach explicitly maps the data to a feature space where it is linearly separable. …The evaluation results shows that the proposed approach provides competing scores with the methods in the available literature. Show more
Keywords: Music/speech, random kitchen sink, feature vector, GTZAN database, S&S database, spectral features
DOI: 10.3233/JIFS-179716
Citation: Journal of Intelligent & Fuzzy Systems, vol. 38, no. 5, pp. 6353-6363, 2020
Authors: Vinayakumar, R. | Soman, K.P. | Poornachandran, Prabaharan | Sachin Kumar, S.
Article Type: Research Article
Abstract: Long Short-term Memory (LSTM) is a sub set of recurrent neural network (RNN) which is specifically used to train to learn long-term temporal dynamics with sequences of arbitrary length. In this paper, long short-term memory (LSTM) architecture is followed for Android malware detection. The data set for evaluation contains real known benign and malware applications from static and dynamic analysis. To achieve acceptable malware detection rates with low computational cost, various LSTM network topologies with several network parameters are used on all extracted features. A stacked LSTM with 32 memory blocks containing one cell each has performed well on detection …of all individual behaviors of malicious applications in comparison to other traditional static machine learning classifier. The architecture quantifies experimental results up to 1000 epochs with learning rate 0.1. This is primarily due to the reason that LSTM has the potential to store long-range dependencies across time-steps and to correlate with successive connection sequences information. The experiment achieved the Android malware detection of 0.939 on dynamic analysis and 0.975 on static analysis on well-known datasets. Show more
Keywords: Android malware detection: static and dynamic analysis, deep learning: recurrent neural network (RNN), Long Short-term Memory (LSTM)
DOI: 10.3233/JIFS-169424
Citation: Journal of Intelligent & Fuzzy Systems, vol. 34, no. 3, pp. 1277-1288, 2018
Authors: Vinayakumar, R. | Soman, K.P. | Poornachandran, Prabaharan | Sachin Kumar, S.
Article Type: Research Article
Abstract: In recent years, domain generation algorithms (DGAs) are the foundational mechanisms for many malware families. Mainly, due to the fact that DGA can generate immense number of pseudo random domain names to associate to a command and control (C2) infrastructures. This paper focuses on to detect and classify the pseudo random domain names without relying on the feature engineering or any other linguistic, contextual or semantics and statistical information by adopting deep learning approaches. A deep learning approach is a complex model of traditional machine learning mechanism that has received renewed interest by solving the long-standing tasks in artificial intelligence …(AI) related to the field of natural language processing, image recognition, speech processing and many others. They have immense capability to extract optimal feature representations by taking input as in the form of raw input texts. To leverage this and to transfer the performance enhancement in aforementioned areas towards characterize, detect and classify the DGA generated domain names to a specific malware family, this paper adopts deep learning mechanisms with a known one million benign domain names from Alexa, OpenDNS and a corpus of malicious domain names generated from 17 DGA malware families in real time for training in character and bigram level and a trained model has been evaluated on the OSNIT data set in real-time. Specifically, to understand the effectiveness of various deep learning mechanisms, we used recurrent neural network (RNN), identity-recurrent neural network (I-RNN), long short-term memory (LSTM), convolution neural network (CNN), and convolutional neural network-long short-term memory (CNN-LSTM) architectures. Additionally, to find out an optimal architecture, experiments are done with various configurations of network parameters and network structures. All experiments run up to 1000 epochs with a learning rate set in the range [0.01-0.5]. Overall, deep learning approaches, particularly family of recurrent neural network and a hybrid network (where the first layer is CNN and a subsequent layer is LSTM) have showed significant performance with a highest detection rate 0.9945 and 0.9879 respectively. The main reason is deep learning approaches have inherent mechanisms to capture hierarchical feature extraction and long range-dependencies in sequence inputs. Show more
Keywords: Domain generation algorithms (DGAs), deep learning mechanisms, recurrent neural network (RNN), identity-recurrent neural network (IRNN), long short-term memory (LSTM), convolution neural network (CNN), convolutional neural network-long short-term memory (CNN-LSTM)
DOI: 10.3233/JIFS-169423
Citation: Journal of Intelligent & Fuzzy Systems, vol. 34, no. 3, pp. 1265-1276, 2018