Research
The Framework of Anomaly-based Android Malware Analysis
Thousands of mobile malware appears every day, most of them targeting smartphones devices with Android as the operating system. The signature-based approach was ineffective in detecting new malware. Anomaly-based malware detection has the ability to detect unknown malware. However, no clear framework that can describe a standard design for developing an anomaly-based framework for malware detection. This research proposed the design framework of anomaly-based malware detection that can be easily adopted by other systems. The detection framework consists of dataset collector, feature processing, detection engine, and evaluation. On the other side, anomaly-based malware detection may produce high false rate alarm. This research also observes how to reduce false positives rates (FPR) and deal with obfuscation techniques. This research proposed features engineering that combines the static and dynamic feature and used multimodal deep learning to maximize the benefit of encompassing combination features. So, it will provide the detection system with high accuracy, low FPR, and robust to obfuscation.
Keyword: detection framework, anomaly-based detection, multimodal deep learning, false positive rate, obfuscation
Automatic Ontology Construction from Unstructured Text Source
Knowledge that is disseminated on the internet at various sources is intended for humans only. In the meantime, knowledge needs to be interpreted and understood not only by humans but also by machines. Information in a format that machines can understand can be used for different purposes, such as: being a knowledge base for reasoning, sharing knowledge between machines, semantic search, information visualization, etc. Ontology learning is a tool that can extract information on a document or web page from unstructured text and then transform it into a knowledge base in a format that machines can understand, namely in the form of ontology. The method consists of several phases, namely: extraction of term, concept, extraction of relation, and evaluation of ontology.
This study explores how the automatic ontology can be developed by extracting knowledge from unstructured text. It also discussed how to improve the quality of the ontology generated in two ways, namely: to conduct a combination of methods at the extraction stage of the concept, and to conduct a double assessment. The YAKE method is used in combination with language-processing techniques at the term extraction stage. DBPedia is used as the basis for extraction at the concept extraction and relating extraction stages. Evaluation is applied in three steps, each after the method of extraction of term, extraction of definition, and extraction of relation. The ontology created through ontology learning is hoped to have good quality in both ways.
Keyword: ontology learning, term extraction, concept extraction, relation extraction, ontology evaluation.
Question Answering System for Students Admission using the Ontology Knowledge Base (Case Study Sumatra Institute of Technology)
In fact , information on welcoming new students and other events that preceded it has been widely available on web pages and brochures. There are complex questions, however, whose responses are not known. However, the use of search engines does not guarantee that users can get correct answers.
This research develops a question answering system related to the admission of new students in order to provide users with adequate and true value information. Question answering system is constructed using natural language processing methods as an interface, and the knowledge base is graph ontology. A query sentence will be processed into several stages to be able to answer a question, namely: preprocessing, language processing, then showing the correct response.
Keyword: question answering system,natural language processing, ontology.
Relief Identification System In Historical Site Using Mobile Device (Borobudur Temple Case Study) - Master Thesis
The great nation built from people who can respects they history and origins. Reliefs at Borobudur temple contained many stories, including the history and origins of this nation. Starting from the life story of the royal, society, and customs at the time of the building of the temple was made, and so on.
This study develops mobile Android software for identification of Borobudur Temple relief image object so that it can help travelers in translating the story and the information contained therein. Feature extraction method used is speeded-Up Robust Feature (SURF) and hierarchical k-means tree nearest-neighbor for identification.
Identification testing of relief images is done by different testing variations, ie angle, distance, orientation change, intensity of the light and wholeness of image input to see the effect on the relief image recognition results. The proposed identification method gives recognition results of 93.30% and the average computation time for 59.55 seconds.
Keyword: Keywords: Borobudur tample, relief, SURF, hierarchical k-means tree.
Get the Borobudur Temple dataset image here.