Speaker: Ferhat Özgür Çatak

Title: Machine Learning for Cyber Security and Secure Model Training

Date/Time: February 20, 2019 / 13:40 - 14:30

Place: FENS 2019

Abstract: With the increase in the number of devices used today, the complexity of malware and cyber attacks against them is also increasing. Cyber attacks are now possible not only between countries but also through our personal phones. For this reason, the issue is now transformed into a personal struggle. In the new generation of communication systems, the number of devices connected to the Internet with the IoT concept increases and the data to be processed within the cyber-space becomes big data. In particular, with the emergence of adversarial machine learning, malware may become easier to bypass security components such as anti-virus, IPS / IDS. To prevent these attacks, static and dynamic analysis methods are not sufficient for new generation malware. This seminar will provide information on tools such as Argus, Scapy, which are used to extract attributes from PCAP files (obtained by using tools such as Wireshark or Tshark) containing malicious network traffic. Ensemble methods or deep learning algorithms used in the development of "malicious traffic detection model" with datasets extracted using the tools will be shared. Another important area of cyber security is malware analysis. The use of LSTM networks in analyzing malware with "metamorphic" properties will be explained, and the creation of CNN models by converting these malware into 8-bit image files will be discussed. Finally, studies on "adversarial machine learning", which will be used in cyber attacks to bypass the security components such as firewall, antivirus, will be shared.

Bio: Dr. Çatak graduated from Eskişehir Osmangazi University, Electrical and Electronics Engineering Department in 2002. In January 2014, he received his Ph.D. degree from Istanbul University, Institute of Science & Engineering, Department of Informatics. He has been working at TÜBİTAK BİLGEM National Electronic and Cryptology Research Institute (UEKAE) between 2011-2014 and at Cyber Security Institute (SGE) since 2014. Since 2016, he has been teaching "Machine Learning for Cyber Security" and "Penetration Testing/Security Auditing" courses at the cyber security engineering graduate program of Istanbul Sehir University and Gebze Technical University. Dr. Çatak is a technical team member at the NATO Science and Technology Organization IST-163 Deep Learning for Cyber Security. His research interests are cyber security, malware analysis, adversarial machine learning, homomorphic encryption algorithms, and topological data analysis.

Contact: Erdinç Öztürk  & Öznur Taştan