Wireless Security

My work is mainly focused on security aspects of wireless networks. In particular, my current research deals with selfish/malicious behaviors at the MAC layer. In a nutshell, my work includes the following:

* Measurement driven, game theoretic modeling of the interactions between a link employing frequency hopping and a jammer. The main application of this model is the performance evaluation of frequency hopping as anti-jamming technique (published in WiOpt 2009).

* Exploitation of bit rate and power control for anti-jamming operations. Based on measurements we obtained from our testbed, we provide guidelines for building an anti-jamming system based on these two physical layer functionalities (published in ACM CoNext 2009 -- arXiv:0906.3038).

* Design and implementation of a gradient descent based, jammer's localization algorithm. The logic of gradient descent minimization can be used in order to design a simple, lightweight and accurate jammer localization algorithm (published in IEEE GlobeCom 2009 - in collaboration with Prof. I.Koutsopoulos from University of Thessaly).

* Identification and detection of an intelligent jamming attack in WLANs. An intelligent jammer can exploit the performance anomaly of 802.11 DCF mode and bring down the aggregate AP's throughput with extremely low power consumption. We propose, implement and evaluate on our testbed a detection scheme based on the packet transmission delay per client, as well as a mitigation scheme based on traffic shaping via packet size tuning (published in SecureComm 2009).

* Detection of selfish users who manipulate their CCA thresholds. This is a protocol compliant, extremely effective, MAC layer selfish behavior. After showing via experiments the effects of this behavior, we have developed and tested a simple, light-weight and efficient detection algorithm (published in INFOCOM 2009 - in collaboration with Dr. Stephan Eidenbenz and Dr. Guanhua Yan from the Los Alamos National Labs).

* Securing the probing functionality of link quality aware routing protocols. Probing functionality is designed with the assumption of benign nodes. Nevertheless, malicious entities can manipulate and deceive legitimate nodes during the computation of the link quality metric. We provide a security mechanism to eliminate this vulnerability for both single and colluding attackers (under submission).

* Dealing with replay attacks (e.g. nodes imposing bogus packets in the network or previously overheard ones). After quantifying the effects of this simple to launch DoS attack, we propose the use of a combination of various security countermeasures to account for the tradeoff between performance and robustness (under submission).

* Building a trust model for wireless network users/nodes/devices. Trust is updated based on the direct interactions with the users as well as with "social gossiping". Social networking aspects are being considered in order to built a robust trust model that can be integrated with higher layer protocols (e.g., routing) and machine learning techniques are being applied. We seek to examine the tradeoff between trustworthy protocol operations and "best-performance" based operations (under preparation).

* Operation provenance is defined as the history of the packet transmissions that have taken place in a network. The availability of hop-level and end-to-end provenance has not been studied before. Clearily, requiring a certain level of provenance introduces a specific hit in the capacity of the network and vice versa (allowing for a specific hit in the network's capacity we can guarantee a specific level of provenance). We build analytical models to identify the factors that affect the availability of provenance and we examine the trade-off between provenance availability and throughput capacity (under preparation).



Next Generation Communications

In addition, I am working on 3G and 4G (MIMO) communications. In particular, I am working on understanding the behavior of 802.11n links via experimentation and applying the knowledge obtained for design of cross layer mechanisms. Furthermore, I am involved in the design and implementation of hybrid wireless network architectures utilizing 3G and WiFi technologies in multihoming settings. In a nutshell, my work up to now includes:

* Examination of the effects of channel bonding on the performance of 802.11n networks. Rules for better exploitation of the MIMO PHY layer from 802.11n are provided (under submission).

* Design of a Multi-User MIMO Successive Interference Cancelation topology control framework. Centralized and distributed schemes are being provided to approximate the optimal topology control (shown to be NP-hard) (published in IEEE INFOCOM 2010).

* Extensive experimentation for (i) characterizing the performance of an 802.11n link in both isolated settings and multi-user settings and (ii) providing guidelines for a correct deployment of an 802.11n testbed (published in WICON 2008).

* Characterization of MIMO links through an accurate but not too complicated abstraction. After understanding the interactions between PHY and higher layers, we design a new abstraction which incorporates PHY layer features not previously considered. We compare our abstraction using benchmark measurements obtained from our 802.11n testbed (to appear in IEEE ICDCS 2010).

* Exploiting MIMO technology for wireless home networking. In particular, we try to understand the features that constitute the gains of 802.11n over legacy systems, and which ones are those that limit its performance. We start with isolated links (utilizing fully controllable environments -- e.g. anechoic chamber, operations at 5 GHz) and we use the knowledge obtained for understanding the more complicated and realistic scenarios of (a) lossy links and (b) multi-user settings. Our work will lead to the design of an adaptive system, which will try to exploit MIMO capabilities under 802.11n to the extend possible (under preparation/part of it will appear at ACM MobiCom '09 poster session - in collaboration with Dr. T.Salonidis and Dr. H.Lundgren from Thomson Research, Paris and Prof. N.Vaidya from University of Illinois Urbana Champaign).

* Computation of the dowlink capacity of a hybrid ad-hoc cellular wireless network. Our analytical and simulation results show that gains are viable under specific parametric regimes (in ACM/IEEE ToN, March 2010).

* Multihoming networks using 3G gateways will become popular within next years due to the plug-n-play and the larger geographical availability of the 3G connections. Economic benefits are viable with connections sharing as well. Careful assignment of connections to gateways in conjunction with fair share of the available bandwidth are some of the challenging task we are looking at (work in progress - in collaboration with Dr. Christos Gkantsidis from the Microsoft Research Cambridge, UK).


Data Fusion algorithms for Network Intrusion Detection

This project involves the study and analysis of data fusion algorithms as the core of Intrusion Detection Systems. Three main algorithms were studied: (i) Baysian inference, (ii) Dempster-Shafer's theory of evidence and (iii) Principal Component Analysis. We implemented and evaluated the performance of these algorithms on real traffic data from attacks recorded at the router of NTUA. Our main findings can be summarized in the following:

Finally, we tried to apply the Eigen Co-Occurence Matrix algorithm on Intrusion Detection Systems. We suggest ways in order for this technique to be used in the future for intrusion detection (part of this work was published at ICNS 2007).