Cyber threats are a showstopper for Internet of Things (IoT) has recently been used at an industrial scale. Network layer attacks on IoT can cause significant disruptions and loss of information. Among such attacks, routing attacks are especially hard to defend against because of the ad-hoc nature of IoT systems and resource constraints of IoT devices. Hence, an efficient approach for detecting and predicting IoT attacks is needed. Systems confidentiality, integrity and availability depends on continuous security and robustness against routing attacks. We propose a deep-learning based machine learning method for detection of routing attacks for IoT. In our study, the Cooja IoT simulator has been utilized for generation of high-fidelity attack data, within IoT networks ranging from 10 to 1000 nodes. We propose a highly scalable, deep-learning based attack detection methodology for detection of IoT routing attacks which are decreased rank, hello-flood and version number modification attacks, with high accuracy and precision. Application of deep learning for cyber-security in IoT requires the availability of substantial IoT attack data and we believe that the IoT attack dataset produced in this work can be utilized for further research.

Özetçe -Günümüzde geleneksel imza tabanlı tespit yöntemleri kullanan anti-virus uygulamaları, metamorfik zararlı yazılımları tespit etmede başarısızdır. Bu nedenle son zamanlarda yapılan tespit ve sınıflandırmaya yönelik çalışmalar, zararlı yazılımların davranışlarını ele almaktadır. Bu çalışma kapsamında, 8 farklı türdeki gerçek zararlı yazılımların API çagrıları kullanılarak, LSTM tabanlı bir sınıflandırma yöntemi geliştirilmiştir. Bu yöntem ile işletim sistemi üzerindeki zararlı yazılım türlerine ait davranışlar modellenmiştir.Anahtar Kelimeler-Metamorfik zararlı yazılımlar, Windows API, derin ögrenme, LSTM.Abstract-Nowadays, anti-virus applications using traditional signature-based detection methods fail to detect metamorphic malware. For this reason, recent studies on the detection and classification of malicious software address the behavior of malware. In this study, an LSTM based classification method was developed by using API calls of 8 different types of real malware. With this method, the behaviors of the malware types on the operating system are modeled.

At the present time, machine learning methods have been becoming popular and the usage areas of these methods have also increased with this popularity. The machine learning methods are expected to increase in the cyber security components like firewalls, antivirus software etc. Nowadays, the use of this type of machine learning methods brings with it various risks. Attackers develop different methods to manipulate different systems, not only cyber security components, but also image detection systems. Therefore, securing machine learning models has become critical. In this paper, we demonstrate a data poisoning attack towards classification method of machine learning models and we also proposed a defense algorithm which makes machine learning models more robust against data poisoning attacks. In this study, we have conducted data poisoning attacks on MNIST, a widely used character detection data set. Using the poisoned MNIST dataset, we built classification models more reliable by using a generative model such as AutoEncoder.

No abstract

No abstract