Early Detection of Network Attacks Using Deep Learning

The Internet has become a prime subject to security attacks and intrusions by attackers. These attacks can lead to system malfunction, network breakdown, data corruption or theft. A network intrusion detection system (IDS) is a tool used for identifying unauthorized and malicious behavior by observing the network traffic. State-of-the-art intrusion detection systems are designed to detect an attack by inspecting the complete information about the attack. This means that an IDS would only be able to detect an attack after it has been executed on the system under attack and might have caused damage to the system. In this paper, we propose an end-to-end early intrusion detection system to prevent network attacks before they could cause any more damage to the system under attack while preventing unforeseen downtime and interruption. We employ a deep neural network-based classifier for attack identification. The network is trained in a supervised manner to extract relevant features from raw network traffic data instead of relying on a manual feature selection process used in most related approaches. Further, we introduce a new metric, called earliness, to evaluate how early our proposed approach detects attacks. We have empirically evaluated our approach on the CICIDS2017 dataset. The results show that our approach performed well and attained an overall 0.803 balanced accuracy.

Evaluating System Identification Methods for Predicting Thermal Dissipation of Heterogeneous SoCs

In this paper we evaluate the use of system identification methods to build a thermal prediction model of heterogeneous SoC platforms that can be used to quickly predict the temperature of different configurations without the need of hardware. Specifically, we focus on modeling approaches that can predict the temperature based on the clock frequency and the utilization percentage of each core. We investigate three methods with respect to their prediction accuracy: a linear state-space identification approach using polynomial regressors, a NARX neural network approach and a recurrent neural network approach configured in an FIR model structure. We evaluate the methods on an Odroid-XU4 board featuring an Exynos 5422 SoC. The results show that the model based on polynomial regressors significantly outperformed the other two models when trained with 1 hour and 6 hours of data.