Download as pdf or txt
Download as pdf or txt
You are on page 1of 27

Detection of Position Falsification Attacks in

VANETs Applying Trust Model and Machine


Learning

Jordan Montenegro 1 Cristhian Iza 1 Mónica Aguilar2

1 Yachay Tech University


2 Universitat Politècnica de Catalunya

17 ACM PE-WASUN 2020

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 1 / 27


Content

1 Introduction

2 Data-Oriented Trust Model

3 Performance Evaluation

4 Conclusions and Future Work

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 2 / 27


Introduction

Content

1 Introduction

2 Data-Oriented Trust Model

3 Performance Evaluation

4 Conclusions and Future Work

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 3 / 27


Introduction

Introduction

Figure 1: VANETs
Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 4 / 27
Introduction

Introduction

Figure 2: Types of trust models

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 5 / 27


Data-Oriented Trust Model

Content

1 Introduction

2 Data-Oriented Trust Model

3 Performance Evaluation

4 Conclusions and Future Work

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 6 / 27


Data-Oriented Trust Model

Direct trust computation


Our approach is a data-oriented model, based in 1 .

Figure 3: Direct trust computation

1
Hasrouny et al. 2019. Trust model for secure group leader-based communications in VANET
Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 7 / 27
Data-Oriented Trust Model

Direct trust computation

The direct trust value is computed as follows:


 1/Np
Np
(i,v) 
Y
DT (i, v) =  αj · mj (1)
j=1

where:
DT (i, v) : Direct trust value of vehicle i computed by vehicle v.
Np : Number of considered parameters; in our case, Np = 2.
αj : Weigh factor for parameter j in our case α1 belongs to speed
parameter = 0.3 and α2 is the weigh factor for received power
coherency = 0.7
mj : Value for parameter j.

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 8 / 27


Data-Oriented Trust Model

Speed normalization. Metric m1

The speed normalization is done as follows:

Figure 4: Speed normalization

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 9 / 27


Data-Oriented Trust Model

Received power coherency. Metric m2

Lf b = 32.4 + 20log(f ) + 20log(d) (2)


Cp = T p − Lf b (3)
Where:
f [M hz] : frequency of the signal (5890 Mhz)
d[Km] : internode distance
Lf b [dB] : Free path loss
Cp : calculated power
T p : transmission power

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 10 / 27


Data-Oriented Trust Model

Received power coherency. Metric m2

Condition to check:
Cp − T v < Rp < Cp (4)
where:
Cp : Calculated power.
Rp : Received power.
T v : Tolerance value (5dBm).
If Equation (4) is accomplished, the value of the corresponding
metric m2 is 10; otherwise, the value of m2 is 1.

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 11 / 27


Performance Evaluation

Content

1 Introduction

2 Data-Oriented Trust Model

3 Performance Evaluation

4 Conclusions and Future Work

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 12 / 27


Performance Evaluation

Simulation
Veins 5.0 2 , Omnet++ 5.5.1 3 and Sumo 1.2.0 4

Figure 5: Simulation map

2
https://veins.car2x.org
3
www.omnetpp.org
4
www.eclipse.org/sumo
Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 13 / 27
Performance Evaluation

Simulation parameters

Minimum power level -110 dBm


Transmission power 20 mW
Transmission range 580m
Bit Rate 6 Mbps
Noise floor -98 dBm
Beacon interval 1 ms
Vehicles’ density 100 vehicles/km2
Road speed limit 15 m/s
Simulation time 15 s
Simulation Area 1.5 km × 1.5 km
MAC Protocol IEEE 802.11p
Vehicular network protocol WAVE
Path loss model Free path model

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 14 / 27


Performance Evaluation

Experimental Setup

Messages x y z
Fixed position 750 750 0
Random position [0,1500] [0,1500] 0
Fixed offset +250 -100 0
Random offset [-300,+300] [-300,+300] 0
Table 1: Types of fake messages

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 15 / 27


Performance Evaluation

Experimental setup

The experiments were divided in the following configurations.

Configuration Direct trust


measures
1) Validation
122 1.1) Well behaved 1000
122 1.2) Misbehaved 1000
2) Half fixed position 1000
3) Half random position 1000
4) Half fixed offset position 1000
5) Half random offset position 1000
Table 2: Experiments

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 16 / 27


Performance Evaluation

Machine learning algorithm used to predict


misbehaving vehicles: K-NN

Figure 6: k-NN algorithm


Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 17 / 27
Performance Evaluation

Machine learning algorithm used to predict


misbehaving vehicles: K-NN 5

k = 101
Training stage:
80% of validation configuration(1600)
Test stage:
Validation configuration (400)
Half fixed position configuration (1000)
Half random position configuration (1000)
Half fixed offset position configuration (1000)
Half random offset position configuration (1000)

5
Silverman et al. 1989. E. Fix and J.L. Hodges(1951): An important contribution to nonparametric discriminant analysis and
density estimation
Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 18 / 27
Performance Evaluation

Evaluation metrics

Accuracy: Number of correct predictions


TP + TN
Accuracy = (5)
TP + FP + FN + TN
Recall: Amount of misbehaved vehicles that were correctly
classified
TP
Recall = T P R = (6)
TP + FN
False Positive Rate: Amount of well behaved vehicles incorrectly
classified as misbehaved
FP
FPR = (7)
TN + FP

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 19 / 27


Performance Evaluation

Evaluation Metrics

Receiver operating characteristic (ROC) curve: A curve plotted


as the true positive rate(TPR) against the false positive rate (FPR).
Area under the curve (AUC): Represents degree or measure of
separability.

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 20 / 27


Performance Evaluation

Simulation Results

Experiment Accuracy Recall


1) Validation 0.958 1.00
2) Half fixed position 0.862 0.747
3) Half random position 0.910 0.889
4) Half fixed offset position 0.912 0.860
5) Half random offset position 0.847 0.779
Table 3: Experiments results

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 21 / 27


Performance Evaluation

Simulation Results
ROC Curve of Experiments

1.0

0.8
True Positive Rate

0.6

0.4

0.2 AUC Half Fixed = 0.85


AUC Half Offset = 0.91
AUC Half Random = 0.91
AUC Half Random Offset= 0.86
0.0 AUC VALIDATION = 0.96
0.0 0.2 0.4 0.6 0.8 1.0
False Positive Rate

Figure 7: ROC curve of experimental configurations

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 22 / 27


Conclusions and Future Work

Content

1 Introduction

2 Data-Oriented Trust Model

3 Performance Evaluation

4 Conclusions and Future Work

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 23 / 27


Conclusions and Future Work

Conclusions

Effectiveness of the proposed model to detect misbehaving nodes


in VANETs.
Scalable model.
Good efficiency of K-NN.

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 24 / 27


Conclusions and Future Work

Future work

Addition of more parameters to the computation of direct trust.


Implementation of complete trust model with public key
infrastructure (PKI).

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 25 / 27


Detection of Position Falsification Attacks in
VANETs Applying Trust Model and Machine
Learning

Jordan Montenegro 1 Cristhian Iza 1 Mónica Aguilar2

1
Yachay Tech University
2
Universitat Politècnica de Catalunya

17 ACM PE-WASUN 2020

Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 26 / 27


References
Hamssa Hasrouny, Abed Ellatif Samhat, Carole Bassil, and Anis
Laouiti. 2019. Trust model for secure group leader-based
communications in VANET.Wireless Networks 25, 8 (2019),
4639–4661. https://doi.org/10.1007/s11276-018-1756-6
Daniel Jiang and Luca Delgrossi. 2008. IEEE 802.11 p: Towards
an internationalstandard for wireless access in vehicular
environments. InVTC Spring 2008-IEEE Vehicular Technology
Conference. IEEE, 2036–2040.
https://doi.org/10.1109/VETECS.2008.458
Bernard W Silverman and M Christopher Jones. 1989. E. Fix and
J.L. Hodges(1951): An important contribution to nonparametric
discriminant analysis and density estimation: Commentary on fix
and hodges (1951).International StatisticalReview/Revue
Internationale de Statistique 57, 3 (1989), 233–238.
https://doi.org/10.2307/1403796
Montenegro, Iza, Aguilar Detection of Position Falsification 17 ACM PE-WASUN 2020 27 / 27

You might also like