VANET. Gilles Guette and Bertrand Ducourthial. IEEE MoVeNet 2007, Pisa. Laboratoire Heudiasyc, UMR CNRS 6599 Université de Technologie de Compiègne

Transcription

1 1 1 out + On the Gilles Guette and Bertrand Ducourthial Laboratoire Heudiasyc, UMR CNRS 6599 Université de Technologie de Compiègne IEEE MoVeNet 2007, Pisa

2 Outlines 2 2 out : hypotheses vs. impact of s 4 5

3 Plan 2 3 out : hypotheses vs. impact of s 4 5

4 out + Network of vehicles mobiles communicating nodes with unstable topology few access to an infrastructure C2C vs. C2I generally highly dynamic ad hoc networks identity preservation Inter-vehicles applications Need for security Solutions 3 4

5 out + Network of vehicles Inter-vehicles applications infrastructure oriented with C2I e.g., gathering information related to traffic conditions vehicle oriented e.g., obstacle detection, position on the road, distance to previous car driver oriented e.g., best way to reach the station, weather forecast, gaz prices... passengers oriented e.g., shared juxe box, touristic informations, games... Need for security Solutions 3 5

6 out + Network of vehicles Inter-vehicles applications Need for security same as in ad hoc networks passive listening routing disturbance specific to depends on the applications traffic jam illusion creation of accidents... easy : just send the right messages Solutions 3 6

7 out + Network of vehicles Inter-vehicles applications Need for security Solutions pessimistic security PKI not easy in sparse connexion to the infrastructure, revocation scheme?... optimistic security combining several sources of information what in case of fake nodes creation? 3 7

8 Plan 3 8 out : hypotheses vs. impact of s 4 5

9 Creation of fake nodes [Douceur et al. 2002] by sending messages with false GPS positions in the transmission area 4 9 out + R F R F R F S F F R F F

10 out + Hypotheses in the litterature Same material or same technical characteristics true if the material comes from a single authority allows to test the resources radio range, delai for several messages... [Newsome et al. 2004] CPU computational puzzles [Douceur et al. 2002] Public key infrastructure available Local checking Continuous propagation model Nodes cooperation Tests and validation scenarios 5 10

11 out + Hypotheses in the litterature Same material or same technical characteristics Public key infrastructure available key unique identifier access to a reliable revocation system several keys for privacy preservation Local checking Continuous propagation model Nodes cooperation Tests and validation scenarios 5 11

12 out + Hypotheses in the litterature Same material or same technical characteristics Public key infrastructure available Local checking there exists a physical mean to associate a sender to a real device (e.g., camera) [Golle et al. 2004] Continuous propagation model Nodes cooperation Tests and validation scenarios 5 12

13 out + Hypotheses in the litterature Same material or same technical characteristics Public key infrastructure available Local checking Continuous propagation model this allows to compute the exact expected reception power depending on the given sender-receiver distance [Xiao et al. 2006] Nodes cooperation Tests and validation scenarios 5 13

14 out + Hypotheses in the litterature Same material or same technical characteristics Public key infrastructure available Local checking Continuous propagation model Nodes cooperation combining local informations [Golle et al. 2004] Tests and validation scenarios 5 14

15 out + Hypotheses in the litterature Same material or same technical characteristics Public key infrastructure available Local checking Continuous propagation model Nodes cooperation Tests and validation scenarios few informations on the scenarios used in a, random positions are not realistic nor representative 5 15

16 Plan 5 16 out : hypotheses vs. impact of s 4 5

17 out + hypotheses vs. impact of s Aim characterize the impact of a number of cheated vehicles characterize the influence of the hypotheses on the impact Method Hypotheses 6 17

18 out + hypotheses vs. impact of s Aim Method computing the geographic area where the succeeds depending on the kind of antenna and the of the sending power P snd Hypotheses 6 18

19 out + Aim Method Hypotheses hypotheses vs. impact of s GPS, coordinates announced in messages measures of the received power sender (er) standard P snd : propagation with tuned P snd : + receiver omnidirectional antenna bidirectional antenna : front/rear 6 19 Sender Receiver omnidirectional antenna bidirectional antenna standard P snd tuned P snd

20 out + How to detect a fake node At the reception, a node compares the measuring reception power and the computed reception power computed with coordinates announced in the messages out nor of P snd P rcv (S, R) = P snd G SR 1 d 2 (S,R) R detects that F S if P rcv (S, R) P rcv (F, R) 7 20 F d max S R More complex with or

21 Plan 7 21 out : hypotheses vs. impact of s 4 out + 5

22 out + without No exact position lines (instead of area) gives the borders of further areas Standard P snd without and omni-directional antenna cheated nodes are on line segment [A, B] (D) A 8 22 F d max S B

23 out + without Standard P snd without and bi-directional antenna and unknown directions the can be detected from everywhere depends on the direction of the receiver : R 1 is cheated, R 2 not. (D) A R 2 F 9 23 d max S R 1 (D R1 ) (D R2 ) B

24 out + without Standard P snd without and bi-directional antenna and same directions cheated nodes are on [A, C] and [D, B] R is cheated if it sees S and F in the same semi-plan. (D) A C (DS) F (D F) direction of all the vehicles d max S D B

25 with real environment unknown upper bound computed with the free space propagation model (factor α) out + Standard P snd with and omni-directional antenna transmission range reduced ; cheated nodes cannot be close to S. C α (D) A C α max F α d max S O α B

26 (DF) On the Sybil Standard P snd with and bi-directional antenna and same directions with R is cheated if it sees S and F in the same semi-plan out + (D) C A α Cmax α α d max S (DS) F direction of all the vehicles OαO α B

27 out + with + + of the sending power increasing factor β for the sending power γ = α β α : factor Standard P snd with + and omni-directional antenna C γ max (D) A C γ Oγ F γ d max S C γ min B

28 (DF) On the Sybil with + Standard P snd with + and bi-directional antenna and same directions out + C γ max (D) A (DS) C γ F direction of all the vehicles Oγ γ d max S C γ min B

29 Plan out : hypotheses vs. impact of s 4 5

30 Rapport On the Sybil When P snd increases the transmission range increases the area of success decreases Sender : impact of the factor out + (D) the er will be detected more easily while his will cheat less vehicles. A Surface de réussite/surface totale O F O O α d max S alpha B

31 out + Receiver : impact of the bidirectionnal antenna When bidirectional antenna are used when the er does not know the directions of the victims, it will be discover bidirectional antenna very useful in urban environments many vehicles with different directions on roads, the Sybil nodes should be placed near the er less impact of the Sybil node creation when using combination of the information in neighborhood 16 31

32 out + Conclusion Summary Realistic hypotheses GPS, measure of the received power Geometric characterization of the success, metric for comparison (area) Attacker : or not Sender : omni- or bidirectional antenna guide for further tests and scenarii design for studies sending power : limited interest bidirectional antenna : urban environment Future work Cooperation in neighborhood