MOBILITY, SECURITY, PRIVACY PRESERVING IN ITS A PEOPLE...

MOBILITY, SECURITY, PRIVACY PRESERVING IN ITS

A PEOPLE CENTRIC APPROACH

Supervisor:

Prof. Dr. Guojun Wang

Department of computer Science

and Technology

Presenter: Muhammad Arif

PhD Student

MS THESIS DEFENSE

7/18/2018

Intelligent Transportation System Muhammad Arif 2 of 36

Introduction (VANETs)

Problem Statement

Progress

Paper (Trusted Communication Scheme Between Vehicles and

Infrastructure Based on Fog Computing)

Paper (Deep Learning with Non-Parametric Regression Model for

Traffic Flow Prediction)

My today’s talk will comprise


Vehicles connected to each others through an ad hoc formation form a wireless

network called “Vehicular Ad Hoc Network”

VANETs are wireless networks where vehicles are both network hosts and routers

They are involved in traffic and safety management. By using V2I and V2Vcomuunicatiobs

A typical VANETs consists of

Road Side Units

Administration and application servers

Location based Service

Proxy Servers

Vehicles

Registration authority

System Model ( VANETS)


VANET

In VANETs scenario we consider 4

entities

1. Traveler

2. Vehicle

3. Infrastructure

4. Attacker

In the coming years, VANETswill performed a significantrole in growing roadsecurity, safety and thetraffic efficiency intransportation systems.


Track me if you can? Query Based Dual Location Privacy in VANETS for V2V and V2I

Communication ( Accepted in Trustcom 2018).

Secure VANETs: Trusted Communication Scheme Between Vehicles and Infrastructure

Based on Fog Computing (Published in SIC Studies in Informatics and Control IF-1.020,

SCI, ESI (Engineering))

Deep Learning with Non-Parametric Regression Model for Traffic Flow Prediction (Accepted

in DASC 2018).

Vehicular trajectory Privacy ( In progress )

A Survey towards Security, Privacy Attacks in VANETS: Communication, Application and

Challenges.

Progress


Secure VANETs: Trusted Communication

Scheme Between Vehicles and

Infrastructure Based on Fog Computing

(Published in SIC IF=1.020)


Overview of the System Architecture Phase I. In this phase vehicle driver initiate the communication

process and generate the messages M. this message M may contain

the sub messages M = m1; m2; :::; mk. After initiating the

communication process the vehicle driver v encrypt the message and

generate key (PK) based on ATB encryption.

Phase II. In this phase, the Fs received the messages M = m1; m2;

:::; mk via different roots. The Fs combined the messages and

decrypt the messages based on the PK received by the vehicle v. All

the Fs perform the same task.

Phase III. In this Phase the Fog anonymizer received the messages

from the Fs node. Anonymize the message based on the

anonymization process, and send to the LBS for the results. The Fog

anonymizer perform the same job for anonymization and de-

anonymization, while sending and receiving the messages from the

LBS.

Phase IV. In the last phase the LBS received the messages from the

Fog anonymizer, understand the communication messages, compile

the desired results, and send back to the Fog anonymizer


FOG Computing Architecture


In fog computing, the algorithms of privacy preserving are run among the fog

nodes and the cloud (LBS).

The running algorithms are resources Prohibited at end devices level, they

usually collect the data for the end devices.

For the privacy preservation at the fog nodes the homomorphic encryption is

used for the preservation of the privacy without the decryption.

For the statistical and aggregation differential privacy is applied to validation

of non exposure of privacy of an arbitrary and conflicting single entrance in

data set.

Privacy in Fog Computing


Fs acts as intermediate tier between the vehicles and Fog anonymizer to

provide the secure communication in VANETs.

So, first of all Fog server Fs decrypts the message Mv with the secret key PK

shared through v. then it transmit it to the Fog Anonymizer.

Fog server Fs confides on wireless networks N to handle the messages

mobility and delivery of the vehicles. Despite all the vehicles receives the

messages, vehicle v is the single vehicle with the secret key PK.

And thus, it is the only vehicle that can decrypted the response and enjoy

with the service, and the others did not have the vehicle PK, so they delete

the messages.

Fog Server


The anonymous inquiry process is generated by the vehicle v .

First define the message M and the privacy preferences k. Then, v engender

the message identifier mid and transmit the message M into the k data

movement producing the set of messages {m1; m2; ::mk}.

The resulting messages are distributed between the neighbors vehicles in

the VANETs

Inquired v encrypts every message mi using (Encryption method ) PK shared

among v and Fog server Fs and the affix mid in plain text to it, that is {mi =

{Epk(mi)||mid}} for every {i = 1; k}.

Anonymous Request


The existence of message id mid in all messages allows the vehicles to

categorize dissimilar sub messages associated to the same message M.

Requester v then randomly chose k - 1 vehicles Vs in his

communications range, and send the messages from {m1 to mk} to

every of them It then send these messages to Fog Server.

Upon accepting the messages mi from the every vehicle in the

communication range then v1 first check the mid. If he is already

acknowledged to send the message with same mid, {mid Sent}, v1

transmit mi to next vehicle v2 in the communications range.

Anonymous Request


Anonymous Request

m1m2 m3m4 m5

m4= v6, v7 and v9

m3= v, v4,v7

m2= v, v1

m1= v

m5= v, v3,


Fog Anonymizer

Location Based

Database Server

Query+

Liocation

Information

Fog

Anonymizer Fog anonymizer that is Responsible on Blurring

the Exact Location and query Information

Query+Cloaked

Spatial Region

Privacy Aware

Query

Processor


Anonymous Response

Mr

Mr

MrMr

Encrypted Queries response Mr are broadcasted to

all the vehicles utilized in Fig. 8(a), that are,

{v, v1, v3, v7, v9}. When v collects the message, can

decrypt it with key PK shared by the Fog server F s.

The auxiliary vehicles deleted this message Mr,

because they do not have the key.


Deep Learning with Non-Parametric Regression Model for Traffic

Flow Prediction

Accepted in DASC 2018


Introduction

For decades, inductive loops have been used to detect vehicles before Intelligent

Transportation Systems (ITSs) [1].

Since the 1990s, the arrival of the ITSs, mainly one of its main elements, has brought

traffic management system (TMS) [2] and the widespread systematic implementations

of the different vehicle recognition technologies on the Interstate Highway and the other

most important forms.

These smooth and extensively deployed sensors, once started on the Internet, start to

continuously produce massive real time traffic flow data, some of which has been

operating for the years.

These data aim at enabling traffic engineers to engage in the real time monitoring of

traffic status and to supervise and get better to the practical competence and the safety

and security of the national road system in a appropriate manner [3].


For the evaluation of the traffic situation across all the travel passages at a certain

points along the highway, each lane tool it is necessary to have one detector.

Detectors posted on a single network are generally accepted as a vehicle detection

station (VDS), which are generally collected and reported together [4].

The data usually include volume (number of the vehicles) and control (the amount of

time) for reporting duration, which is commonly 30 seconds.

Some parameters of the traffic, vehicle density, speed, traveling time, can be

accurately find from the traffic data.

Introduction


The traffic status information shows it is very necessary for the many application in

ITS, which includes dynamic traffic allocation.

Researchers and practitioners have immediately realized that the benefits of the ITS

cannot be fully maximized with unrecognized traffic flow parameters in the advance,

specifically, predictions [5].

ITMs can only work in an interactive behavior. Meanwhile, accurate predictions of

traffic conditions will result in timely, proactive, and effective identification of and

solution to problems.

Introduction


This study mainly aims at performing traffic flow forecasts for atypical traffic

conditions through a deep learning and parametric regression model.

The predictive performance of our study has been compared with

that of the other three models:

Amodel SAE proposed by Yisheng [33] ,

BP-neural network (Nnet), and OL-SVR proposed by Manoel Castro [28].

Introduction


Methodology


What is this

unit doing?

Deep Learning


The new way to train multi-layer NNs…



Train this layer first




then this layer




then this layer

then this layer




then this layer

then this layer

then this layer




then this layer

then this layer

then this layerfinally this layer



EACH of the (non-output) layers is

trained to be an auto-encoder

Basically, it is forced to learn good features that

describe what comes from the previous layer


an auto-encoder is trained, with an absolutely standard

weight-adjustment algorithm to reproduce the input


an auto-encoder is trained, with an absolutely standard weight-

adjustment algorithm to reproduce the input

By making this happen with (many) fewer units than the inputs, this

forces the ‘hidden layer’ units to become good feature detectors


intermediate layers are each trained to be auto

encoders (or similar)


Final layer trained to predict class based on

outputs from previous layers


Training Algorithm

Train the first layer as an auto encoder by minimizing the objective function with the

training sets as the input.

Train the second layer as an auto encoder taking the first layer’s output as the input.

Iterate as in 2) for the desired number of layers.

Use the output of the last layer as the input for the prediction layer, and initialize its

parameters randomly or by supervised training.

Fine-tune the parameters of all layers with the BP method in a supervised way.


The input variables include the flow rate V(t− 15 min,k− 2), V(t− 15 min,k− 1). …

V(t min,k− 1), and V(t min,k),

While the output variable is V(t+ 15 min,k),

With t representing the current time and k representing the link section in the

network;

While k− 1 denotes the nearer upstream link section and k− 2 the further one

Input Parameters


Parametric regression model Non-Parametric regressions models are considered as dynamic clustering models.

This technique tries to recognize the sets of previous cases with the inputs or situation values

alike to the systems state during prediction.

They are dynamic in that they define a set of similar previous situations (or neighborhood)

around the state of the current inputs, rather than limiting the number of clusters prior to

prediction time.

Nonparametric regression is a category of regression analysis in which the predictor does not

take a predetermined form but is constructed according to information derived from the data.

https://en.wikipedia.org/wiki/Regression_analysis


Evaluation

To evaluate the proposed models effectiveness, we use five performance indexes, such as(MAE) the mean absolute error, (MRE) mean relative error, and(RMSE) the RMS error, which are defined as where is the

observed traffic flow and is the predicted traffic flow.

We also use absolute percent error APE, and mean

absolute percent error MAPE and RC rate for the

evaluation of our system.

If the MAPE and APE values are lower its mean that the

efficiency of the proposed system is higher and vice

versa.


Results


Results

Prediction performance measured and compared with different state of the art methods. of the comparison betweendifferent techniques and tried to find the APE error.

The efficiency will be more as compared to the minimum value of APE error, furthermore Fig.4 is the bestexplanation of this context.

Some techniques values are very near to our proposed method but due to increase in time the traffic congestionwill increased the accuracy of the other techniques become poor, as compare to our technique .that shows that ourtechnique performs far better than previous techniques


Results

The (MAPE) mean average percent error is shown in the Fig. In this

Fig. we calculated MAPE error, which is calculated by seven vehicle

detectors stations. MAPE error is determinant of the degree of

accuracy, if the MAPE error is higher the accuracy will be higher and

vice versa.

Fig. also represents the OL-SVR method drawback in terms of

congested traffic pattern. When we increase the traffic vehicle

detective stations the other techniques also perform poor but our

technique suitably perform better w.r.t traffic congestion.

This shows high values in different time periods of SAE, OL-SVR

and neural network models. it is clearly predicted that our proposed

method have the lowest MAPE values, its mean that it is most

suitable for the traffic flow prediction in urban areas.

They also show that our proposed technique shows better

performance then SAI, BP-Nnet, and the OL-SVR on different data

collected from the different vehicle detector stations


Results


RC rate is directly proportional to the other parameters. From Fig. (a, b, c) the

comparison of MRE, RMSE, MAE and RC rate is shown, that our technique will outperform

the other techniques because RC rate is lower than other techniques w.r.t MRE, RMSE, MAE.

Our method speaks perfectly the best performance of all the other methods in assessment.

Here, to evaluate statistical side performance, we show the error distributions of models

under a vertical rate of 30 percent. MAE and the RMSE focus on the average distance

between traffic and the real data, while distance can be signifies either by absolute error.

Thus, we only need to be worry about the absolute error allocations and relative error. Fig. 6.

(d) Shows the performance on different traffic data that are collected from different detector

stations on week and non-week days.

Results


Results


In Fig. we compare the congested roads w.r.t time, in which time range from 0 to 2000

seconds. With 30 routes. Clearly understandable from figure that with time increase

in congestion the others techniques did not provide the satisfactory work over the purposed

method.

In Fig. (b; c; d) we compared the mean travel time vehicle arrivals and mean

road converge rate w.r.t to time. From these three figures it is understandable that our method

perform better then the other techniques because our method is most suitable for congested

traffic in urban area.

Fig.7 also shows the results of all traffic management methods, with an average of over 1000

times, with a time limit of 200 seconds. The three methods (SAE, BP-NNet, OL-SVR) did

not give satisfactory results to over the proposed method.

Apart from this, our proposed method gives good results on average with the other three

algorithms in terms of four metrics, and also other methods are not designed to adapt in a

synchronized way under the same framework.

Results


Data Collection

There are a variety of vendors for the different virtualization

types by which cloud environments can be built

http://pems.dot.ca.gov/

http://tris.highwaysengland.co.uk/detail/trafficflowdata

http://pems.dot.ca.gov/

http://tris.highwaysengland.co.uk/detail/trafficflowdata


Thank You !


[1] J. Huang and H.-S. Tan, “Error analysis and performance evaluation of a future-trajectory-based cooperative collision warning

system,” IEEE Trans. Intell. Transp. Syst., vol. 10, no. 1, pp. 175–180, Mar. 2009.

[2] J. Kianfar and P. Edara, “Optimizing freeway traffic sensor locations by clustering Global-Positioning-System-derived speed

patterns,” IEEE Trans. Intell. Transp. Syst., vol. 11, no. 3, pp. 738–747, Sep. 2010.

[3] D. Chen, Y.-S. Fu, B. Cai, and Y.-X. Yuan, “Modeling and algorithms of GPS data reduction for the Qinghaitibet Railway,” IEEE

Trans. Intell. Transp. Syst., vol. 11, no. 3, pp. 753–758, Sep. 2010.

[4] D. Schleischer, L. M. Bergasa, M. Ocaña, R. Barea, and M. E. López, “Real-time hierarchical outdoor slam based on stereovision and

GPS fusion,” IEEE Trans. Intell. Transp. Syst., vol. 10, no. 3, pp. 440–452, Sep. 2009.

[5] J.-I. Meguro, T. Murata, J.-I. Takiguchi, Y. Amano, and T. Hashizume, “GPS multipath mitigation for urban area using

omnidirectional infrared camera,” IEEE Trans. Intell. Transp. Syst., vol. 10, no. 1, pp. 22–30, Mar. 2009.

[6] H. G. Jung, Y. H. Cho, P. J. Yoon, and J. Kim, “Scanning laser-radarbased target position designation for parking aid system,” IEEE

Trans. Intell. Transp. Syst., vol. 9, no. 3, pp. 406–424, Sep. 2008.

[7] S. Gidel, P. Checchin, C. Blanc, T. Chateau, and L. Trassoudaine, “Pedestrian detection and tracking in an urban environment using a

multilayer laser scanner,” IEEE Trans. Intell. Transp. Syst., vol. 11, no. 3, pp. 579–588, Sep. 2010.

[8] P. Zito, H. Chen, and M. C. Bell, “Predicting real-time roadside CO and NO2 concentrations using

neural networks,” IEEE Trans. Intell. Transp. Syst., vol. 9, no. 3, pp. 514–522, Sep. 2008.

[9] K. Sohn and K. Hwang, “Space-based passing time estimation on a freeway using cell phones as

traffic probes,” IEEE Trans. Intell. Transp. Syst., vol. 9, no. 3, pp. 559–568, Sep. 2008.

[10] F. Calabrese, M. Colonna, P. Lovisolo, D. Parata, and C. Ratti, “Real-time urban monitoring using

cell phones: A case study in Rome,” IEEE Trans. Intell. Transp. Syst., vol. 12, no. 1, pp. 141–151, Mar.

2011.

References


[11] T. Gandhi and M. Trivedi, “Vehicle surround capture: Survey of techniques and a novel vehicle blind spots,” IEEE Trans. Intell.

Transp. Syst., vol. 7, no. 3, pp. 293–308, Sep. 2006.

[12] T. Gandhi, R. Chang, and M. M. Trivedi, “Video and seismic sensorbased structural health monitoring: Framework, algorithms, and

implementation,” IEEE Trans. Intell. Transp. Syst., vol. 8, no. 2, pp. 169–180, Jun. 2007.

[13] T. Ehlgen, T. Pajdla, and D. Ammon, “Eliminating blind spots for assisted driving,” IEEE Trans. Intell. Transp. Syst., vol. 9, no. 4,

pp. 657– 665, Dec. 2008.

[14] V. Agarwal, N. V. Murali, and C. Chandramouli, “A cost-effective ultrasonic sensor-based driver-assistance system for congested

traffic conditions,” IEEE Trans. Intell. Transp. Syst., vol. 10, no. 3, pp. 486–498, Sep. 2009.

[15] S. Jwa, U. Özgüner, and Z. Tang, “Information-theoretic data registration for UAV-based sensing,” IEEE Trans. Intell. Transp. Syst.,

vol. 9, no. 1, pp. 5–15, Mar. 2008.

[16] X. Gong and X. Liu, “A data-mining-based algorithm for traffic network flow forecasting,” in Proc. Int. Conf. Integr. Knowl.

Intensive Multiagent Syst., 2003, pp. 243–248.

[17] H. Jula, M. Dessouky, and P. A. Ioannou, “Real-time estimation of travel times along the arcs and arrival times at the nodes of

dynamic stochastic networks,” IEEE Trans. Intell. Transp. Syst., vol. 9, no. 1, pp. 97–110, Mar. 2008.

[18] O. Linda and M. Manic, “Online spatiotemporal risk assessment for intelligent transportation systems,” IEEE Trans. Intell. Transp.

Syst., vol. 12, no. 1, pp. 194–200, Mar. 2011.

[19] D. Vasquez, T. Fraichard, and C. Laugier, “Incremental learning of statistical motion patterns with growing hiddenMarkov models,”

IEEE Trans. Intell. Transp. Syst., vol. 10, no. 3, pp. 403–416, Sep. 2009.

[20] H. Veeraraghavan and N. P. Papanikolopoulos, “Learning to recognize video-based spatiotemporal events,” IEEE Trans. Intell.

Transp. Syst., vol. 10, no. 4, pp. 628–638, Dec. 2009.

[21] P. Angkititrakul, R. Terashima, and T. Wakita, “On the use of stochastic driver behavior model in lane departure warning,” IEEE

Trans. Intell. Transp. Syst., vol. 12, no. 1, pp. 174–183, Mar. 2011.

References (Cont..)

MOBILITY, SECURITY, PRIVACY PRESERVING IN ITS A PEOPLE...

Documents

Transcript of MOBILITY, SECURITY, PRIVACY PRESERVING IN ITS A PEOPLE...