Student Seminars: Kickoff

Transcription

1 Seminars Student Seminars: Kickoff Walid Saad Durham 447

2 Seminars Fall Logistics Weekly meetings in SEB 135 SEB 125 used 10/24, 11/07, and 12/05 Contacts: SaiDhiraj Amuru, Harpreet Dhillon, Walid Saad For the students, by the students So what am I doing here? Due to short notice, no student has stepped up Who will participate? The goal is to have most speakers be students Faculty will be attending sporadically 2

3 Seminars What are the goals of these seminars? Sleeping hour for grad students? A form of torture? Debate group : engaging discussions between groups Integrating new students by giving them insights on how to identify and formulate a research problem, how to choose a technique, etc. Practice prelims, defense, etc. Free pizza (outside)! 3

4 Additional Goals Have talks that discuss new directions in wireless or other areas we re engaged in Talks/tutorials that discuss how to use certain tools For example: how to program a USRP? How to use NS3 for wireless simulations? We will solicit such questions from the students Distinguished Speakers from inside and outside VT We will strive to get at least one distinguished speaker per year from outside VT Build a culture of research in our students community => Sustainable only via active students participation 4

5 Best Practices The purpose of these meetings is not to overly defend your research but rather get feedback from fellow students and explain to them what you re doing Strongly suggest to explain a basic problem you re working on, while clarifying methodology, assumptions made, etc. You know this but. use more figures and illustrations to clarify your ideas However, we need sufficient technical details in these seminars so that new students can learn from the rest 5

6 Best Practices Avoid overly texty slides Concise bullets preferred Avoid weird fonts And weird colors Make eye contact with your audience! Prepare for questions, if no questions, you can start the debate Be enthusiastic, not bored! 6

7 Today s Talk On the Physical Layer Security of Backscatter Systems

8 Outline Introduction to RFID systems and the backscatter Motivation and key features Backscatter communication System model Basic communication model What happens in the presence of eavesdroppers? Physical layer security of backscatter systems Introduction and motivation Basic model Main analysis Simulation results Conclusions and future outlook 8

9 RFID Systems A Radio Frequency Identification (RFID) system is composed of two key components RFID tag: electronic device that can store and transmit data to a reader in a contactless manner RFID reader: Transceiver that can read and communicate with RFID tags 9

10 RFID Systems Three types of RFID tags Passive: cheapest does not include any power source, harvests energy from the reader Semi-Passive: can include some basic sotrage elements as well as a possible small battery to power the circuitry (but not for TX) Active: has transmission capabilities basically a small sensor Two main technologies Near-field communication (NFC): operate at very short distances, e.g., few centimeters, over HF frequencies (i.e., MHz). Ultra high frequency (UHF) RFIDs: operate at UHF frequencies (typically 433 MHz), can be used at longer ranges (10 meters is common but recent research is looking at larger distances) Of interest in this talk 10

11 RFID Applications 11

12 Basic RFID Communication Interrogate and Power Tag Responds with data, by modulating the CW Reader transmits a continuous wave (CW) carrier signal Induces an RF voltage across the tag antenna that is converted to DC by a power harvesting circuit to power the tag s circuitry and demodulate the received command Tag transmits back its stored information by controlling the amount of backscatter of the impinging downlink signal by varying the impedance (mis)-match of the antenna front-end Role of the reader in energizing the tag 12

13 RFID Backscatter: Challenges Tags must remain cheap and passive tags must harvest energy Limitations on circuit design. Design of innovative RFIDs is a key challenge. The uplink tag-to-reader communication has a strictly limited bandwidth due to limited backscatter power and tag collisions How to improve the uplink communication? Crucial to allow higher distances, better rate, and new applications! Security is probably the most critical issue Fact: Classical cryptography hard to implement on RFID tags Current solutions (lightweight cryptography) are non-scalable Security solutions with little complexity => Physical layer security! 13

14 System Model Basic UHF RFID system with one reader and one tag The signal backscattered by a tag k toward the receive antenna of a reader i is given by (Griffin et al., IEEE Trans. Antenna and Propagation, Feb 2008): Noise Tagreader channel Readertag channel Reader s TX signal Tag s reflection coefficient (useful signal) 14

15 Presence of Eavesdroppers A nearby eavesdropper can tap into the communication Two main critics of PHY security But cryptography works fine in cellular systems! But not in RFID You assume eaves. location known RFIDs work on small distances! Does the backscatter constitute the ideal PHY security application? Let s see Tag Responds with data/id Interrogate and power Tag Passive tag requires the reader s continuous wave to power up! 15

16 Key Backscatter Features Eavesdropping in a backscatter system is mostly useful at close distances Due to low-power, low-cost nature of the system as well as the reliance on passive tags Eavesdroppers are mainly interested in the uplink signal/tag data So they must focus their effort on the tag, not the reader The downlink reader-tag signal is also received by the eavesdropper during backscatter A feature to exploit for PHY security 16

17 Signal at the Eavesdropper The signal backscattered by a tag toward the receive antenna of a reader as received at the eavesdropper Tag-eave channel Readertag channel Tag s reflection coefficient (useful signal) Reader continuous CW transmission, Reader-eave channel Noise In a classical system, the eavesdropper knows the CW x and, thus, the second term can be cancelled and does not hinder eavesdropping 17

18 Using Artificial Noise Observation: the backscatter channel provides a mean for reader-induced interference, due to the nature of the communication A random noise signal added to the CW transmitted by the readers will be detrimental to the eavesdroppers Can improve the physical layer security of the system While this has been used in PHY security for conventional systems, it often requires MIMO, in RFID, no need for MIMO! Easily implementable, with little changes at the tag 18

19 Using Artificial Noise Each reader transmits x+zwith z being a random narrowband artificial noise known to the reader With noise injection, the reader s signal becomes: And the eavesdropper s received signal: z and z are the same signal, but received by the eavesdropper over different paths and at possibly different time instants Power of the received noise from z will naturally be much larger than that of the backscattered noise 19

20 PHY Security: Basics C d 1 I can hear User 1! Notion of secrecy capacity C e 1 Maximum rate sent from a wireless node to its destination in the presence of eavesdroppers Here, secrecy capacity of user 1 : C 1 = (C d 1 -C e 1) + For RFID, we use the secrecy capacity as a theoretical performance metric to assess how artificial noise can improve secrecy Practical RFID-tailored notions is an interesting direction 20

21 RFID with Artificial Noise Approximation of secrecy capacity of the backscatter communication between reader and tag: The received powers follow the Friis model We assume worst-case scenario at eavesdropper, i.e., noise is 0, hereinafter. 21

22 Positive Secrecy Key question: For this basic model, what is the condition for positive secrecy? Let s look at two baseline cases Case 1: all terminals have omnidirectional antennas Positive secrecy is challenging with eaves. close to the tag. Case 2: any antennas at the eavesdropper Let s look at antenna impact numerically 22

23 Case 2 Even for highly directional antenna, positive secrecy is possible with reasonable noise power 23

24 Back to Case 1 Even for an eavesdropper placed at 10 cm of the tag, positive secrecy is possible with reasonable noise power 24

25 Optimal Power Allocation Reader has a finite transmit power budget that must be split between noise and main CW How to allocate the reader s finite power between noise and real signal? Analytical closed-form solution can be derived easily (simple optimization problem) 25

26 Simulation results (1) Positive secrecy Achievable even for small distances 26

27 Simulation results (2) Positive secrecy Achievable even for highly directive antennas 27

28 Simulation results (3) Only a small noise injection is sufficient for positive secrecy 28

29 The Multi-Reader Case What if we have multiple readers? Key tradeoff: more power for noise, worse SINR at eavesdroppers, but Less power for the useful signal Coupling in actions: It can lead the other readers to change their power allocation, due to the spreading interference New dimension to the problem Adding noise can protect against eavesdroppers but the power of each others noise will impact our performance We now have a problem where decisions between readers are interdependent 29

30 The Multi-Reader Case What are the techniques we can now use? We can still use an optimization problem for the entire system Advantage: ensure system-wide optimality Disadvantage: could require additional overhead and explicit control of all readers (may not be possible in practice) Game theory Why? Interdependence in decisions between the readers Advantage: allows individual optimization per reader (kind of like we did before) Disadvantage: could yield an inefficient point and may still require some form of information at each reader. 30

31 RFID Game Formulation What kind of game? Noncooperative, because we assume no explicit communication between readers We formulate the problem as a strategic, noncooperative game with The readers being the players The action of each reader being the fraction of power allotted to the noise and the useful signal The utility being the sum of secrecy rates at all tags, i.e., 31

32 Game Solution Definition: A Nash equilibrium is a strategy profile s* with the property that no player i can do better by choosing a strategy different from s*, given that every other player j i. For each player i with payoff function u i, we have: Does the Nash equilibrium exist? Not always, multiplicity, efficiency, algorithmic aspects Algorithmic solution via best response Not necessarily the best solution, but can give you insights on a problem 32

33 Best-Response Solution Initial State: -No artificial noise being used -- Discovery of neighboring readers Sequentially, Each reader decides on its optimal power distribution, under the currently observed network state Stage I Best Response Dynamics, until finding a Nash equilibrium - Readers send out the interrogate commands using the derived powers/secrecy - The tags respond using backscatter Stage II Actual data acquisition 33

34 Simulation results (1) Classical approach = Regular channel With no noise Case with 6 Eavesdroppers: Performance advantage particularly when the number Of eavesroppers > Number of readers 34

35 Simulation results (2) Reasonable Number of iterations 35

36 Summary The physical layer security of backscatter RFID systems is still a largely unchartered territory Most natural application for PHY security, due to infeasibility of cryptography and the need for cost-effective, practical solutions Our preliminary results show that a simple solution can lead to significant gains, at least theoretically Future work Advanced signal processing and information theoretic analysis More system-level analysis, multi-reader/tag/eavesdropper Theoretical analysis on performance of RFID PHY security Innovative optimization techniques, beyond basic artificial noise 36

37 Finally. Thank You Questions? 37