Alva Erwin

Analysis of Machine learning Techniques Used in Behavior-Based Malware Detection

The increase of malware that are exploiting the Internet daily has become a serious threat. The manual heuristic inspection of malware analysis is no longer considered effective and efficient compared against the high spreading rate of malware. Hence, automated behavior-based malware detection using machine learning techniques is considered a profound solution. The behavior of each malware on an emulated (sandbox) environment will be automatically analyzed and will generate behavior reports. These reports will be preprocessed into sparse vector models for further machine learning (classification). The classifiers used in this research are k-Nearest Neighbors (kNN), Naïve Bayes, J48 Decision Tree, Support Vector Machine (SVM), and Multilayer Perceptron Neural Network (MlP). Based on the analysis of the tests and experimental results of all the 5 classifiers, the overall best performance was achieved by J48 decision tree with a recall of 95.9%, a false positive rate of 2.4%, a precision of 97.3%, and an accuracy of 96.8%. In summary, it can be concluded that a proof-of-concept based on automatic behavior-based malware analysis and the use of machine learning techniques could detect malware quite effectively and efficiently.

Automatic mood classification of Indonesian tweets using linguistic approach

Research concerning Twitter mining becomes an interesting research topic recently. It is proven by numerous number of published paper related with this topic. This research is intended to develop a prototype system for classifying Indonesian language tweets. The prototype includes preprocessing step, main information retrieval and classification system. This research proposes a system that uses grammatical rule for retrieving main information from the tweet, and then classifies the information to the suitable mood space. The classification algorithm, which is used, is lexicon based classifier. The proposed classification system has 53.67% accuracy for classifying tweets into 12 mood spaces and 75% accuracy for classifying tweets into 4 mood spaces. As the comparison, the same dataset is also classified using SVM and Naïve Bayes.

Processing performance on Apache Pig, Apache Hive and MySQL cluster

MySQL Cluster is a famous clustered database that is used to store and manipulate data. The problem with MySQL Cluster is that as the data grows larger, the time required to process the data increases and additional resources may be needed. With Hadoop and Hive and Pig, processing time can be faster than MySQL Cluster. In this paper, three data testers with the same data model will run simple queries and to find out at how many rows Hive or Pig is faster than MySQL Cluster. The data model taken from GroupLens Research Project [12] showed a result that Hive is the most appropriate for this data model in a low-cost hardware environment.

Automated document classification for news article in Bahasa Indonesia based on term frequency inverse document frequency (TF-IDF) approach

The exponential growth of the data may lead us to the information explosion era, an era where most of the data cannot be managed easily. Text mining study is believed to prevent the world from entering that era. One of the text mining studies that may prevent the explosion era is text classification. It is a way to classify articles into several predefined categories. In this research, the classifier implements TF-IDF algorithm. TF-IDF is an algorithm that counts the word weight by considering frequency of the word (TF) and in how many files the word can be found (IDF). Since the IDF could see the in how many files a term can be found, it can control the weight of each word. When a word can be found in so many files, it will be considered as an unimportant word. TF-IDF has been proven to create a classifier that could classify news articles in Bahasa Indonesia in a high accuracy; 98.3%.

Indonesian president candidates 2014 sentiment analysis by using Twitter data

The purpose of this research is to find the opinion on Twitter about the 2014 president candidates and find the correlation between the opinion on Twitter and on digital newspaper. To perform this, tweets are extracted. Some tweets will be labelled president candidates name and the positive and negative sentiment for the training set. A training will be conducted to test whether the training set is enough to perform classification or not. The next step is to calculate the sentiment results and compare to the results from digital newspaper by using a web-based application called Tirto. Deep analysing conducted to analyse the relation between the issues on Twitter and on digital newspaper.

Analysis of educational institution DNS network traffic for insider threats

The Internet is a medium for people to communicate with each other. Individuals and/or organizations are faced with increased security threats on the Internet. Many organizations prioritize on handling external security threats over internal security threats and for this reason, internal security threats are often missed or worst ignored. Domain Name System (DNS) is one of major Internet services that resolve users request on domain name to an IP address. Since all of the user query to domain name utilize DNS to resolve the domain name or vice versa, including malicious intended users query. Thus, DNS is a great source of information for detecting potential insider threat to detect unknown insider threats. This research aims to detect insider threats using DNS based features and these potential insider threats are clustered based on the DNS traffic features. Machine learning algorithms are used to cluster the DNS traffic under investigation. Our research shows that suspected clusters of DNS traffic contain insider threats in the organizations and the most frequent suspect of insider threats are botnet, categorized as misuse in insider threat classification. Some clusters could be suspicious indicating insider threats and other cluster is also a benign cluster but potentially an abnormal traffic.

Data mining approach for user profile generation on advertisement serving

Advertisement serving on website is a prosperous business with huge market and millions of dollar prospect. By placing right advertisement at right time and place to right people, advertiser can increase their revenue by huge margin. The question is how advertiser and broker can push the right advertisement to the right user. User profiling can be used to analyze users behavior and predict what kind of advertisement should be served to the website user. Data mining approach can be harnessed to help with user profiling process. With data mining technique, users trace can be used as data source for behavior analysis. This research is used to do user profiling based on their browsing history stored on proxy server. Their browsing history serves as the basis of content crawling for content analysis using Multinomial Naïve Bayes classifier based text classification. The result of profiling then will be used as the basis for serving advertisement to user. The result of content analysis is validated by asking users preferences and comparing it with profile generated by classifier engine.

Automatic multi-document summarization for Indonesian documents using hybrid abstractive-extractive summarization technique

This paper discusses the development of multi-document summarization for Indonesian documents by using hybrid abstractive-extractive summarization approach. Multi-document summarization is a technology that able to summarize multiple documents and present them in one summary. The method used in this research, hybrid abstractive-extractive summarization technique, is a summarization technique that is the combination of WordNet based text summarization (abstractive technique) and title word based text summarization (extractive technique). After an experiment with LSA as the comparison method, this research method successfully generated a well-compressed and readable summary with a fast processing time.

Automatic text summarization based on semantic analysis approach for documents in Indonesian language

Research about text summarization has been quite an interesting topic over the years, proven by numerous number of papers related with discussion of their studies such as approaches, challenges and trends. This papers goal is to define a measurement for text summarization using Semantic Analysis Approach for Documents in Indonesian language. The applied measurement requires Indonesian version of WordNet which had been implemented roughly. The main idea of semantic analysis is to obtain the similarity between sentences by calculating the vector values of each sentence with the title. The need ofWordNet is to define the depth of each word as being computed for word similarity. Combining all required formulas and calculations, a compact and precise summarization is produced without depriving the gist information of certain document.

News recommendation in Indonesian language based on user click behavior

Recommendation system has been proposed for years as the solution of information era problem. This research strives to develop an intelligent recommendation system based on user click behavior on news websites. We extracted frequent itemsets and association rules from the web server log of a news website, performed a pre-computation of similarity between news articles, and then proposed a three-level recommendation system: based on association rule discovery, news articles on the same category, and similarity between news articles. By combining collaborative filtering approach and content-based filtering, experiment results show that the technique produces reliable news recommendation.