Setiadi Yazid

A survey for handwritten signature verification

Signature verification is the process used to recognize an individuals handwritten signature. Signature verification can be divided into two main areas depending on the data acquisition method, off-line and on-line signature verification. In this paper we attempt to survey the signature verification based on three categories. First, judging from how to get the data signature which is off-line and on-line verification. Second, based on the technique used, that is rule-based approach, neural networks, hidden Markov model and support vector machine. Third, based on preprocessing and feature extraction, which is thinning and line segmentation. Based on the survey, it was concluded that any method of verification has advantages and disadvantages. However, if viewed from the ease of implementation and performance, using neural networks or hidden Markov models are the right choice. Depending on the data acquisition method, on-line verification is recommended to use than off-line verification.

Identification of Gene Expression Linked to Malignancy of Human Colorectal Carcinoma using Restricted Boltzmann Machines

Learning hidden information or pattern on gene expression data to uncover an underlying molecular features is called gene expression profiling. To perform gene expression profiling, an unsupervised machine learning method can be employed. In this paper, Gaussian RBM is proposed to obtain the optimal number of clusters and their members on human colorectal cancer dataset provided by Muro. Gaussian RBM forms two large numbers of genes clusters and one smaller cluster which has several tumour-classifier genes as its members. The two large numbers of genes clusters formed by Gaussian RBM succeed in showing a significant correlation with the existence of tumour and distant metastasis but they show no significant correlation with lymph node metastasis existence. The smaller number of genes clusters gives a statistically significant result in clustering patients into two groups.

Cancer subtype identification using deep learning approach

In this paper, a framework using deep learning approach is proposed to identify two subtypes of human colorectal carcinoma cancer. The identification process uses information from gene expression and clinical data which is obtained from data integration process. One of deep learning architecture, multimodal Deep Boltzmann Machines (DBM) is used for data integration process. The joint representation gene expression and clinical is later used as Restricted Boltzmann Machines (RBM) input for cancer subtype identification. Kaplan Meier survival analysis is employed to evaluate the identification result. The curves on survival plot obtained from Kaplan Meier analysis are tested using three statistic tests to ensure that there is a significant difference between those curves. According to Log Rank, Generalized Wilcoxon and Tarone-Ware, the two groups of patients with different cancer subtypes identified using the proposed framework are significantly different.

Design DDoS attack detector using NTOPNG

Distributed Denial of Service (DDoS) is one kind of attacks using multiple computers. An attacker would act as a fake service requester that drains resources in computer target. This makes the target cannot serve the real request service. Thus we need to develop DDoS detector system. The proposed system consists of traffic capture, packet analyzer, and packet displayer. The system utilizes Ntopng as main traffic analyzer. Detector system has to meet good standard in accuracy, sensitivity, and reliability. We evaluate the system using one of dangerous DDoS tool named Slowloris. The system can detect attacks and provide alerts to detector user. The system also can process all incoming packets with a small margin of error (0.76%).

Wavelet denoising and fractal feature selection for classifying simulated earthquake signal from mobile phone accelerometer

This work is an initial study of the research that aims to help people by giving an information about the earthquake while it happens eventhough the phone is not connected to the internet. In this research, we identify the pattern of the simulated earthquake signal from the mobile phone accelerometer via machine learning. Before the data is processed into the classifier, static windowing and denoising was done to boost up the accuracy. Another fractal features are extracted from the pre-denoised data, which are the box counting dimension feature and the Hurst coefficient. The purpose of doing static windowing is to obtain more features so that we can have many potential useful attribute candidates as possible. Denoising with symlet wavelet is done to remove the noises which can worsen the classification accuracy. The classification is done using support vector machine and multilayer perceptron classifier with the accuracy of 81% and 82.15%, respectively.

Certificate policy analysis and formulation of the government public key infrastructure using SSM: A case study of Indonesia State Cryptography Agency (Lemsaneg)

This research was motivated by the expanding scope of Public Key Infrastructure (PKI) implementation held by Indonesia State Cryptography Agency (Lembaga Sandi Negara — Lemsaneg). Previously it was only within the scope of the Government Electronic Procurement Service, and now it must be expanded to meet the needs of government agencies services in general. With the expanding scope of the PKI, the Certificate Policy must also be reevaluated and a new policy is formulated. In this study, the new Certificate Policy (CP) is formulated using Soft Systems Methodology process. The proposed Certificate Policy is based on Levels of Assurance (LoA) to accommodate the need of different government bodies. The certificate profile analysis follows the Certificate Policy Framework of IETF RFC 3647. NIST Internal Report 7924 has been selected as the template for the Certificate Policy document due to its closeness to PKI implementation requirements in Lemsaneg and the latest information system security demands.

Multimodal Deep Boltzmann Machines for feature selection on gene expression data

In this paper, multimodal Deep Boltzmann Machines (DBM) is employed to learn important genes (biomarkers) on gene expression data from human carcinoma colorectal. The learning process involves gene expression data and several patient phenotypes such as lymph node and distant metastasis occurrence. The proposed framework in this paper uses multimodal DBM to train records with metastasis occurrence. Later, the trained model is tested using records with no metastasis occurrence. After that, Mean Squared Error (MSE) is measured from the reconstructed and the original gene expression data. Genes are ranked based on the MSE value. The first gene has the highest MSE value. After that, k-means clustering is performed using various number of genes. Features that give the highest purity index are considered as the important genes. The important genes obtained from the proposed framework and two sample t-test are being compared. From the accuracy of metastasis classification, the proposed framework gives higher results compared to the top genes from two sample t-test.

Comparative study of lightweight secure multiroute communication system in low cost wireless sensor network for CO 2 monitoring

Wireless Sensor Network (WSN) is a system used to conduct a remote monitoring in a wide monitoring area. It has a sensor node — a sampling point — which communicate each other to passing their data to central node for recapitulation or transmit it to data center. Because of that, communication system is a crucial thing for WSN. However, WSN may be deployed in a environment that far from ideal condition. Placed in an unattended area with far distance between nodes, WSN is very vulnerable with security threats. To overcome that, the good combination between communication protocol and encryption algorithm for WSN is needed to gather an accurate and representative data with high transmission speed. This research focused on finding those combination for our own-made low-cost sensor node for CO2 monitoring. In this research, two routing protocols (AODV and TARP) and several encryption algorithms (AES, ChaCha, and Speck) tested to determine which combination is give the best result. As the result, combination between routing protocol AODV and encryption algorithm Speck give the best result in the term of performance.

Can smartphones be used to detect an earthquake? Using a machine learning approach to identify an earthquake event

The possibility of using smart phone accelerometer to detect earthquake is investigated in this research. Experiments are designed to learn the pattern of an earthquake signal recorded from smart phones accelerometer. The signal is processed using N-gram modeling as feature extractor for machine learning. For the classifier, this study use Naïve Bayes, Multi-Layer Perceptron (MLP), and Random Forest. Our result shows that the best classification accuracy is achieved by Random Forest method. Its accuracy reached 93.15%. It can be concluded that smart phones can be used as an earthquake detector.