Prof. Dr. Dogan Kesdogan
Lehrstuhl für Wirtschaftsinformatik IV
Universität Regensburg
D - 93040 Regensburg
Tel.: 0941-943-5901
Telefax: 0941-943-5902
Electronic Mail: kesdogan@ur.de
REPORT
on
Ph.D. thesis
' On selected privacy and security issues in wireless sensor networks1
Submitted by
Mr. Jiří Kůr
General Evaluation
Mr. Kur investigates in his Ph.D. thesis selective privacy and security issues in
Wireless Sensor Networks (WSN). On the one hand the work can be characterized by
an extended version of a cumulative work. Each Chapter is based on published own
papers. On the other hand, all chapters have a „Ieitmotif\ i.e. cost-efficiency,
privacy&security and especially intrusion detection. Thus, with this leitmotif Mr. Kur
manages to build a common and strong story starting with the analysis of attacker's
possibilities and the conflict between privacy and Intrusion Detection Systems (IDS).
After this analysis (or as a result of this analysis) the constructive part starts with the
development of privacy enhanced IDS, adaptive security architecture and improved
key predistribution algorithms. The work is an important contribution to applied
security&privacy research.
Mr. Kur has published 12 papers (see Appendix A). It is an impressive list of
international journals and magazines but also international conferences and
workshops. Consequently, Mr. Kur can base his Chapters of his thesis to several high
quality publications. However, due to the investigation of many aspects of the
security&privacy of WSN an in-depth analysis on one aspect can not be given by the
very nature of the matter. It gives more a comprehensive analysis on several selected
topics that are closely related.
Structure
Mr. Kur subdivides his work in seven chapters. It starts and ends with the obligatory
introduction and conclusion chapters. At the first glance the work is a monograph but
looking closer it has also elements of a cumulative thesis. The reason for that is that
Mr. Kur has investigated several aspects of WSN problem and presents it more or
less in autonomous chapters. Consequently, the chapters follow mostly a classical
JiH Kur:On selected privacy and security issues in wireless sensor networks 2
structure with introduction, related works and contribution parts. However, the
chapters are also linked to each other. The linkage between the chapters is given by
the general problem orientation with the following aspects: How to estimate and
fairly compare the performance of security mechanisms, what are the fundamental
contradictions between IDS and Privacy and how to bring them together. Since, many
research areas have to be considered the overall structure is adequate to handle the
investigated scientific problem.
Content
Chapter 1 represents a very short introduction into the WSN area. It introduces the
general problem statement, contributions and the general structure of the thesis.
Security research starts usually with the question of defining a concise attacker
model. The general focus is here on global eavesdropper (passive attack) with certain
malicious active actions (local active attacker). It is essential to know the concrete
abilities of the attacker in this attack space. Otherwise, designing security without
concise attacker model the security design could simply fail (or simply unpractical).
Therefore it is important to search and explore the attackers vector space. After a
general introduction (e.g. related works) Mr. Kur presents his approach to explore the
abilities of the attacker, i.e. uses evolutionary algorithms to construct automated
attacks. Obviously, this search can not be exhaustive since genetic algorithms are an
optimization strategy to accelerate the search. However, the given reasons for using
genetic algorithms are vague (solution space are between 100 to one million) and
need to be better justified.
In Chapter 3 he investigates the principle structure between IDS and privacy.
Network privacy aims to hide the actions of a particular node and the goal of IDS is
to identify all actions of a node. Hence, realizing both approaches are inherently
contradictory and they have to be technically mitigated. Obviously, it is hard (even
impossible) to find optimal general solutions. Mr. Kur identifies twelve problems that
have to be addressed and afterwards four approaches to mitigate the problems.
Clearly, to enable cooperation between the two approaches one has to investigate
special aspects of techniques and has to harmonize them, e.g. probing message versus
dummy message. However, if IDS is considered as the „internal" attacker then the
situation changes. Nevertheless, this could be addressed by distributing sensitive data
and responsibility among the sensor nodes. And this is one of the main ideas that is
suggested in the work. Taken as a whole this chapter is very important since it
addresses the general problem of addressing privacy and IDS.
In Chapter 4 a location privacy friendly intrusion detection is presented. The work is
based on the link layer security scheme SNEP. Privacy is provided by simply
periodic collection of information. The IDS Problem is to detect malicious nodes that
Jiří Kůr:On selected privacy and security issues in wireless sensor networks 3
modify the forwarded messages. Mr. Kur suggests a protocol where every node
becomes a watchdog and controls the predecessor and the successor node. He has
also implemented the sketched solution and can therefore present real measurements.
The suggested IDS works well if single node or pairs of nodes are malicious.
However, how about if we assume more successive nodes are malicious? Since Mr.
Kur has implemented the protocol he could experiment and measure many aspects,
e.g. increased traffic load, drop rate and false alarm rate. However, all these aspects
are left for future work.
Chapter 5 starts with an analysis of adaptive security management. The argument for
dynamic adaptive security level management is to save resources. Of course the more
level one has the better adjustment can be achieved. However, the management cost
increases with the number of levels. After presenting the related works Mr. Kur gives
a solution for a specific application scenario. In this frame he considers five attacker
models that result in five security levels. He claims that the internal, active and local
attacker model is the most likely to appear in reality. However, this claim is not
explained. Mr. Kur presents his findings in a systematic and clear way and presents
the additional cost of each level. The final subsections deal with the question of
management of the levels and dynamic pseudonym changes.
Chapter 6 presents two improvements of random key distribution. Mr. Kur shows
here a stringent analytic approach. After presenting his extensions a mathematical
analysis is given. He thereupon sketches computational results (results of network
simulator) that verify the analytical results. However, I wonder why he has not
compared these results directly by using the figures. Chapter 7 concludes with the
obligatory open questions part.
The work of Mr. Kur can be extended in many ways. He presents an analysis on
several selected topics that are closely related. One can identify many further steps to
investigate in each topic but this shows also how fruitful his work is. Considering the
breadth of the selected topics and the purpose to give a practical general solution the
work is well balanced in depth and breadth.
The work is excellently written and explained. I recommend the Faculty of
Informatics, Masaryk University to accept the submitted PhD thesis. My overall
rating is:
Sum ma cum laude (1.0).
Regensburg, February 02,2014
— ^ ( — \ ' L
(Prof. Dr. Dogan Kesdogan)