| Title | Graph-Based Malware Detection Using Opcode Sequences |
| Publication Type | Conference Paper |
| Year of Publication | 2021 |
| Authors | Gülmez, Sibel, Sogukpinar, Ibrahim |
| Conference Name | 2021 9th International Symposium on Digital Forensics and Security (ISDFS) |
| Date Published | jun |
| Keywords | Encryption, feature extraction, graph-based detection, Hardware, Histograms, Human Behavior, Image edge detection, Malware, malware analysis, Malware Analysis and Graph Theory, malware detection, Metrics, opcode analysis, Packed Malware, Predictive Metrics, privacy, pubcrawl, resilience, Resiliency, static analysis |
| Abstract | The impact of malware grows for IT (information technology) systems day by day. The number, the complexity, and the cost of them increase rapidly. While researchers are developing new and better detection algorithms, attackers are also evolving malware to fail the current detection techniques. Therefore malware detection becomes one of the most challenging tasks in cyber security. To increase the performance of the detection techniques, researchers benefit from different approaches. But some of them might cost a lot both in time and hardware resources. This situation puts forward fast and cheap detection methods. In this context, static analysis provides these utilities but it is important to keep detection accuracy high while reducing resource consumption. Opcodes (operational codes) are commonly used in static analysis but sometimes feature extraction from opcodes might be difficult since an opcode sequence might have a great length. Furthermore, most of the malware developers use obfuscation and encryption techniques to avoid detection methods based on static analysis. This kind of malware is called packed malware and according to common belief, packed malware should be either unpacked or analyzed dynamically in order to detect them. In this study, a graph-based malware detection method has been proposed to overcome these problems. The proposed method relies on obtaining the opcode graph of every executable file in the dataset and using them for future extraction. In this way, the proposed method reaches up to 98% detection accuracy. In addition to the accuracy rate, the proposed method makes it possible to detect packed malware without the need for unpacking or dynamic analysis. |
| DOI | 10.1109/ISDFS52919.2021.9486386 |
| Citation Key | gulmez_graph-based_2021 |
Graph-Based Malware Detection Using Opcode Sequences