Elektronik ve Haberleşme Mühendisliği Bölümü
Permanent URI for this collectionhttps://hdl.handle.net/20.500.12416/179
Browse
Browsing Elektronik ve Haberleşme Mühendisliği Bölümü by Author "Akar, Arif"
Now showing 1 - 1 of 1
- Results Per Page
- Sort Options
Item Citation Count: Akar, Arif (2017). Machine learning based anomaly detection technique for in-vehicle networks / Araç içi ağlar için makine öğrenmesi tabanlı anomali tespit tekniği. Yayımlanmış yüksek lisans tezi. Ankara: Çankaya Üniversitesi Fen Bilimleri Enstitüsü.Machine learning based anomaly detection technique for in-vehicle networks(Çankaya Üniversitesi, 2017) Akar, Arif; Çankaya Üniveristesi, Fen Bilimleri Enstitüsü, Elektronik ve Haberleşme Mühendisliği Anabilim DalıThe automotive industry faces a revolution by connecting vehicles to the communication infrastructure in the scope of intelligent transportation systems (ITS). The idea of internet of things (IoT) entering the automotive domain raises much skepticism about security and privacy issues. The information received from and sent to vehicles bears considerable risks for all components in the transportation system. Commonly, the IT industry uses firewall devices to filter communication in both receiving and transmitting directions that require heavy maintenance personnel support and instant configuration changes. Considering the mobility of vehicles and the light-weight nature of in-vehicle networks, firewalls require too many resources and miss automated decision making. Intrusion detection systems (IDS) are widely used in traditional IT networks and try to close gaps resulting from stateful firewalls. This thesis proposes the In-Vehicle Anomaly Detection Engine (IVADE) as an anomaly based intrusion detection algorithm for in-vehicle controller area network (CAN) applications using machine learning methods. The algorithm aims at detecting malicious manipulations of vehicle mobility data (such as position, speed, direction) which are exchanged in the form of Cooperative Awareness Messages on vehicle to vehicle (V2V) and vehicle to infrastructure (V2I) networks. The functionality of IVADE is validated by simulations of a Lane Keeping Assistance system that is implemented on a CAN bus together with the electronic control units (ECUs) for signal measurement and control computations. The relevant features for applying machine learning in IVADE are derived from received CAN message fields, supported with automotive domain-specific knowledge of the dynamic system behavior and trained with Decision Trees. The obtained simulation results indicate that IVADE successfully detects anomalies in in-vehicle applications and hence supports safety-critical functions.