View of A Novel Authentication Framework with Conditional Privacy Preservation and Non-Repudiation for Fog-Vanet

(1)

8353 http://annalsofrscb.ro

A Novel Authentication Framework with Conditional Privacy Preservation and Non-Repudiation for Fog-Vanet

Dr.V. Sharmila¹, K. Jamuna², K. Jeevitha³, I. Kalam⁴, Dr.V. Vennila⁵

1Professor, Department of Computer Science and Engineering, K.S.R College of Engineering, Tiruchengode, Tamil Nadu, India.

2B.E Student, Department of Computer Science and Engineering, K.S.R College of Engineering, Tiruchengode, Tamil Nadu, India.

5Associate Professor, Department of Computer Science and Engineering, K.S.R College of Engineering, Tiruchengode, Tamil Nadu, India.

ABSTRACT

The joining between conveyed figuring and vehicular uniquely delegated associations, explicitly, vehicular fogs (VCs), has become an enormous investigation domain. Profound learning is essential for a more extensive group of AI strategies dependent on fake neural organizations with portrayal learning. Quite possibly the most fascinating regions of exploration is the examination of street traffic. This incorporates vehicle way following, way forecast, smart vehicles, clog location and some more. The majority of the exploration that has been never really gridlock utilized vehicular specially appointed organization (VANET) yet of late information mining approach has been applied. In spite of the fact that the vast majority of the proposed work has effectively recognized gridlock, it is intricate to think of a successful instrument that fuses identification, control and forecast of repetitive and non- intermittent gridlocks across the board framework. This venture takes a gander at how information mining contrasts and VANET in performing street gridlock identification, control and expectation. By planning the computation capacity of a cloud laborer with that of vehicles nowadays, we propose a Deep Neural Network (DNN) part to normally and adroitly perceive wild vehicles dependent on AI counts. To help a significant mindful disturbing structure, a three-level system designing is made from existing vehicular associations, which involves vehicles, street side units (RSU) and a cloud laborer. Furthermore, given the way that most vehicles can be trusted to drive safely, to moreover diminish the transmission stack of the DNN segment, we plan our plot so every vehicle simply moves the data of driving moves with indiscreet potential to RSUs. In view of the gathered data, the cloud specialist overall rates every vehicle's driving display by using Support Vector Machine (SVM) and Decision Tree (DT) decision tree AI models. We finally execute the proposed DNN part into a standard traffic test framework, Simulation of Urban Mobility (SUMO), for appraisal.

Entertainment results address that our protected disturbing structure can decisively distinguish insane vehicles and effectively give advantageous aler.

KEYWORDS

VANET, Machine Learning, SVM.

(2)

Introduction

Machine learning (ML) is the study of PC calculations that improve consequently through experience. AI calculations construct a model dependent on example information, known as

"preparing information", to settle on expectations or choices without being expressly modified to do as such. AI calculations are utilized in a wide assortment of uses, for example, email separating and PC vision, where it is troublesome or impractical to create customary calculations to play out the required errands. A subset of AI is firmly identified with computational insights, which centers around making forecasts utilizing PCs; however not all AI is factual learning. The investigation of numerical advancement conveys strategies, hypothesis and application spaces to the field of AI. Information mining is a connected field of study, zeroing in on exploratory information examination through unaided learning. In its application across business issues, AI is additionally alluded to as prescient examination.

Deep Neural Network

Profound learning (otherwise called profound organized learning) is important for a more extensive group of AI strategies dependent on counterfeit neural organizations with portrayal learning. Learning can be directed, semi-administered or solo. Profound learning models, for example, profound neural organizations, profound conviction organizations, intermittent neural organizations and convolutional neural organizations have been applied to fields including PC vision, machine vision, discourse acknowledgment, regular language handling, sound acknowledgment, informal community sifting, machine interpretation, bioinformatics, drug plan, clinical picture examination, material assessment and tabletop game projects, where they have delivered results similar to and now and again astounding human master execution. Fake neural organizations (ANNs) were motivated by data preparing and dispersed correspondence hubs in natural frameworks. ANNs have different contrasts from organic cerebrums.

Sumo Simulation

Traffic reenactments encourage the assessment of foundation changes just as strategy changes before they are actualized out and about. For instance, the viability of natural zones or traffic signal control calculations can be tried and upgraded in a reenactment prior to being sent in the genuine world. SUMO (Simulation of Urban MObility) is a free, open, minuscule and constant street traffic reproduction suite intended to deal with enormous street organizations. It permits demonstrating of multi-purpose traffic frameworks including street vehicles, public vehicle and people on foot. SUMO incorporates an abundance of supporting apparatuses, which handle undertakings, for example, course discovering, perception, network import and emanation estimation. SUMO can be improved with custom models and gives different APIs to distantly control the reproduction.

(3)

Reckless Driving

To more readily comprehend the results of wild driving, we initially portray the qualities of careless driving, and afterward hypothetically investigate its outcomes through a contextual analysis. A. Careless Driving Performance Although further developed security systems are used in present day autos these days than at any other time, the quantity of driving fatalities is as yet disturbing high. Indeed, any driving conduct dismissing the wellbeing of others can be viewed as foolish driving.

Reckless Driving Defensive Alarming System

Given the extreme results of wild driving, we propose an AI based protective disturbing framework to proactively recognize and caution foolish vehicles. In particular, we plan a Deep Neural Network (DNN) component to naturally recognize crazy vehicles, whose location results are shipped off the drivers close by. These capacities are upheld by a three-level framework design. In the accompanying, we initially present the DNN instrument, and afterward present our framework engineering followed by the flagging interaction.

Related Work

Qian Luo, Yurui Cao et.al has proposed with the developing number of exploration concentrates on self-ruling driving, a great deal of new advances for self-driving vehicles are likewise continually arising. While altogether advancing the improvement of self-ruling vehicles, these arising advances likewise have possible weaknesses, which may raise some nonnegligible security dangers. Escape clauses existing in limitation and route advancements, which are the critical components of course arranging and route, could be used by enemies to control self- sufficient driving route by commandeering important vehicles, products, or even objective characters. Toward this end, we give in this article an essential guide summing up the security dangers and assault plans, and we propose countermeasures explicitly for confinement and route advancements in self-ruling vehicles. From that point forward, we present a novel course mocking assault that could likewise be actualized against future self-sufficient driving route. With such execution, we show that successfully, courses of focused drivers can be usurped and thus deceive drivers to take an off-base course, pass by a vindictive spot, and even be followed continuously.

Jiadai Wang, et.al has proposed the advancement of cell phones with improving correspondence and perceptual abilities has achieved a multiplication of various complex and calculation escalated portable applications. Cell phones with restricted assets face more serious limit limitations than any time in recent memory. As another idea of organization engineering and an augmentation of distributed computing, Mobile Edge Computing (MEC) is by all accounts a promising answer for address this arising difficulty. Nonetheless, MEC likewise has a few impediments, for example, the significant expense of framework arrangement and upkeep, just as the extreme pressing factor that the mind boggling and mutative edge figuring climate brings to MEC workers.

(4)

Lei Zhao, Jiadai Wang, et.al has proposed The idea of keen city has been thriving dependent on the prosperous improvement of different trend setting innovations: versatile edge processing (MEC), super thick systems administration, and programming characterized organizing.

Notwithstanding, it turns out to be progressively confounded to configuration steering procedures to meet the severe and truly changing organization necessities because of the powerful dissemination of the group in various areas of shrewd urban communities. To mitigate the organization blockage and equilibrium the organization load for supporting keen city administrations with sensational inconsistencies, we plan a profound fortification learning-based savvy directing calculation to make the dispersed figuring and correspondence framework completely reasonable while at the same time fulfilling the dormancy requirements of administration demands from the group. Other than the proposed calculation, broad mathematical outcomes are likewise introduced to approve its viability. In this article, we have proposed the DRLS defeating calculation to insightfully deal with numerous ser - bad habit solicitations of group dispersion in various shrewd city areas.

Gulbadan Sikander et.al has proposed Driver exhaustion has been credited to car crashes;

consequently, weakness related auto collisions have a higher casualty rate and cause more harm to the environmental factors contrasted and mishaps where the drivers are ready. As of late, many auto organizations have introduced driver help advances in vehicles for driver help. Outsider organizations are additionally fabricating exhaustion discovery gadgets; nonetheless, much exploration is as yet needed for development. In the field of driver weakness recognition, consistent examination is being performed and a few articles propose promising outcomes in compelled conditions, still a lot of progress is required.

Bo Li1, Junjie Yan et.al has proposed Visual article following has been a key theme as of late and numerous profound learning based trackers have accomplished cutting edge execution on different benchmarks. In any case, a large portion of these trackers can scarcely get top execution with ongoing velocity. In this paper, we propose the Siamese area proposition organization (Siamese-RPN) which is start to finish prepared disconnected with huge scope picture sets. In particular, it comprises of Siamese subnetwork for highlight extraction and locale proposition subnetwork including the characterization branch and relapse branch. In the deduction stage, the proposed structure is formed as a neighborhood one-shot location task.

V. Vennila, A. Rajiv Kannan has introduced parallel linguistic fuzzy rule with canopy MapReduce (LFR-CM) framework. In this framework canopy MapReduce function is used to classify big data information sharing in the cloud with higher classification accuracy and lesser time consumption.

P. Balamurugan, M. Shyamala Devi, V. Sharmila has introduced the optimized methods for securing data (OMSD) which is for secure data transmission with trust based weights.

P. Balamurugan, M. Shyamala Devi, V. Sharmila has proposed Score-based data gathering algorithm which provides a significant solution to maximize the network lifetime as well as minimum delay per round of data gathering.

(5)

V. Vennila, A. Rajiv Kannan proposed Discritized Support Vector Classification and Prediction (DSV-CP) model for sharing information in the cloud environment by using efficient Big Data computation.

V. Sharmila, G. Tholkappia Arasu, P. Balamurugan has proposed a non-class element based iterative clustering approach. Weight calculation are used for selection classes.

P. Balamurugan, T. Ravichandran, V. Sharmila proposed Grade- Based Data Gathering (GBDG) algorithm for minimizing wireless sensor networks energy consumption.

V. Sharmila, P. Balamurugan, V. Vennila, S. Savitha has proposed a data verification scheme. In which malicious data packets are identified.[21]

V. Vennila, A. Rajiv Kannan has proposed Parallel Symmetric Matrix-based Predictive Bayes Classifier (PSM-PBC) model is developed for efficient Big Data computation and information sharing in Cloud environment.

Proposed Methodology

The proposed request to give precise foolish driving ready, this segment centers around examining the driving conduct rating model dependent on AI calculations. In the accompanying, we first outline the driving exhibition dataset, and afterward present the driving execution rating model. This undertaking takes a gander at how information mining contrasts and VANET in performing street gridlock recognition, control and forecast. By planning the figuring capacity of a cloud laborer with that of vehicles nowadays, we propose a Deep Neural Network (DNN) part to normally and keenly perceive wild vehicles dependent on AI estimations. To help a serious mindful disturbing system, a three-level structure designing is made from existing vehicular associations, which contains vehicles, street side units (RSU) and a cloud specialist. Perform in a way that is better than VANET in attempting to anticipate blockage location occasions in light of the capacity to utilize verifiable information to perform prescient examination. Plus, given the way that most vehicles can be trusted to drive safely, to moreover diminish the transmission load of the DNN segment, we plan our plot so every vehicle simply moves the data of driving moves with imprudent potential to RSUs. In light of the gathered data, the cloud laborer overall rates every vehicle's driving presentation by using Support Vector Machine (SVM) and Decision Tree (DT) decision tree AI models.

Pre-processing

The information mining approach is in this manner selected versus VANET in order to utilize huge recorded datasets for the motivations behind anticipating intermittent clog occasions. To gather direction information containing boundaries, for example, scope, longitude and timestamp.

The dataset should show blockage occasions. To perform preprocessing on the direction information with the goal that blunders and missing qualities are wiped out or limited. To spatially and transiently group directions to distinguishes vehicles inside a similar space and going simultaneously. For each bunch, and given the timestamp of every single direction point,

(6)

figure the distances between two direction focuses and decide its speed utilizing the distance/time.

RSU Base Traffic Congestion Detection

The information gathered from sensors across the streets for extraction of examples that were utilized for forecast of gridlocks. First and foremost information preprocessing on the gathered information for example cleaning for mistakes, absent or unreasonable qualities is performed.

Besides three methods Deep Neural Network (DNN) using Support Vector Machine (SVM) and Decision Tree (DT) decision tree AI models were relatively actualized on a similar information test. SVM and choice tree calculation makes a choice tree from a quality of the information which is chosen as a root. For each conceivable estimation of the characteristic, a branch is shaped and the cycle is rehashed until the last hub. DNN is an interconnection of neurons that performs calculations on information from various data sources. The calculation gains even from numerous sources of info that perhaps obscure. The Nearest Neighbor is a grouping calculation that arranges tests based on the distance. The aftereffects of the three strategies demonstrated that Decision Tree (DT) has the best precision rate.

Intelligent Transportation Cloud Server

Canny transportation mists could offer Types of assistance, for example, independence, portability, choice help and the standard improvement Environment for traffic the executives techniques, etc. With portable specialist innovation, a metropolitan traffic the executives framework dependent on Agent-Based Distributed and Adaptive Platforms for Transportation Systems (ADAPTS) is both attainable and viable. Notwithstanding, the huge scope utilization of versatile specialists will prompt the rise of an unpredictable, ground-breaking association layer that requires tremendous processing and force assets. To manage this issue, the creators propose a model metropolitan traffic the board framework utilizing multi specialist based keen traffic mists.

Assume a direction for ID1, V_ID 1 is (t1x1y1, t2x2y2,…..tnxnyn) where t is the timestamp, X is scope and Y is longitude for the vehicle direction from guide 1 toward n. The time taken by the direction from guide 1 toward n is given by tn-t1.In this manner the speed of the direction can be determined utilizing,

Speed =

Vehicular Location Tracking

With the paces acquired for all vehicle directions in each bunch, the normal group speed will be determined. The identification of clog will be finished by looking at the bunch speeds got, to a speed edge that is doled out dependent out and about limit. The model will isolate clog from customary stoppages, for example, traffic signal stoppages or picking and dropping travelers by contrasting the span of the recognized occasion to a length edge time.

The model will at that point record the space (organizes) and (time span) for each recognized

(7)

blockage occasion and actualize Jaccard recipe (1) and (2) to decide comparative clog occasions when new information is gotten in order to anticipate intermittent clog occasions.

SVM based Traffic Prediction

Since the preparation of SVM includes a curved advancement issue, the ideal arrangement is a worldwide ideal The upper bound of the speculation blunder is free of issue dimensionality To rate a vehicle's driving exhibition, the info information of SVM and choice tree models is as h21- dimensional highlights, name. In particular, the 21-dimensional highlights are gotten from the standardized driving conduct highlights, including the time span (1 measurement), the level of the distance hole (isolating the general reach into 10 measurements) and the level of the relative speed (partitioning the general reach into 10 measurements). In the accompanying, the SVM and choice tree models are introduced, individually In SVM models, the planning depends on a piece work, which has numerous sorts as the internal item between the planned sets of focuses in the component space. Since piece determination is fundamental to get acceptable grouping results.

Decision based Traffic Prediction

For choice tree model the nonparametric nature and simple understanding make it mainstream in an assortment of use fields. Its bit of leeway over numerous different models is the viability to build arrangements through sectioning an informational index into more modest and more homogeneous groups. A choice tree is a stream graph like tree structure, which infers a characterization choice through a succession of tests along a way of hubs. The root hub contains the whole dataset X = {x1, x,.., xN}. The tree develops through parting information at the hubs, where the active parts of a hub compare to the divided information subsets. At the point when the information of a hub can't be splitted any further, it turns into a terminal hub showing an order choice, i.e., the ordinary or foolish rating choice of a vehicle. In our choice tree model based classifier, we consider three well known determination calculations to locate the best split indicator Deep Neural Network (DNN), curve test, and communication test. The standard DNN chooses the split indicator that expands the split-model addition over all potential parts, everything being equal; the curve test chooses the split indicator that limits the p-estimation of Chi-square trial of freedom among indicators and the reaction; and comparably collaboration test dependent on the count measure of curve test thinks about the cooperation between each pair of indicator and reaction. We assess the exhibition of these three calculations in our assessment part.

To introduce a solid presentation, we use the standard DNN calculation as the choice calculation to depict the choice tree model preparing in the ensuing turn of events. The standard of tree developing advance is to recursively parcel target factors so the information in the relative hubs are consistently cleaner than that in the parent hub. At the point when a preparation information enters the root hub, a test is performed to look for all potential parts dependent on a parting standard that gauges the parting quality.

Performance Evaluations

We assess the adequacy of our cautious alarming framework, where the dataset readiness is initially depicted, trailed by the exhibition assessment including two basic viewpoints. In

(8)

particular, we initially assess the precision of our driving exhibition rating, just as the presentation examination between our AI based technique and the ordinary measurable based method. In the accompanying, we assess the exactness of the three rating strategies, i.e., SVM, choice tree and factual driving execution rating techniques. In the DNN instrument, we use various detecting gadgets to completely portray distinctive wild driving moves, where the AI based driving execution rating models are acquainted due with the trouble of breaking down the multi-modular detecting information. Since the result of an accident comes from the topographically vicinity between vehicles, a wild vehicle with high accident likelihood instinctively drives nearer to neighbor vehicles.

Along these lines, rather than the multi-modular detecting highlights, we ought to assess the accessibility of a solitary component, i.e., the distance hole, to rate a vehicle's "foolish" level. By utilizing the single distance hole include, the traditional factual strategy can quantify a vehicle's driving presentation through contrasting the vehicle's measurable distance hole, i.e., mean distance hole, with the experimental limit got from recently distinguished foolish vehicles with low calculation intricacy.

Experimental Setup

Experimental results demonstrate the Exploratory outcomes exhibit the viability of our Thus, because of the profoundly powerful nature of driving conditions, the AI based driving execution rating models by utilizing multi-modular detecting information are importantly vital. By contrasting the four measurements, we can see that the accuracy and review measurements are with the least and the most noteworthy qualities, separately. As indicated by their relating definitions, their assessment results confirm that albeit a couple of bogus alerts exist in the framework, virtually all wild vehicles can be recognized, which can be acknowledged for the security thought in our guarded cautioning framework. Likewise, to completely use our dataset for a more precise correlation, we direct a 10-crease cross-approval for our two AI based rating models, where the deficiency of the SVM and choice tree models are 0.0708 and 0.0606, individually. Hence, these two AI based rating models can give precise rating to distinguish crazy vehicles, and the choice tree model performs better in many conditions. Hence, we use the quantity of detailed careless driving moves to quantify the length of the checking time. We can see that both the two AI based rating models perform better under a bigger number of moves, which implies the more drawn out checking time the more exact rating results are. Also, inside a short checking time, i.e., a couple of detailed careless driving moves, the exactness of our rating is around 85%, which will be improved to around 90% when the quantity of the observed crazy driving occasions For certain territories with generally high accident likelihood, for example, the bar territory, a foolish driving alarm might be sent ahead of time inside a couple of announced information to improve the neighborhood driving wellbeing. What's more, because of the increment of exactness with more revealed crazy driving moves, the rating results ought to be ideal refreshed to end the bogus alert and distinguish the genuine foolish vehicles.

Table 1. Accuracy Evaluation

Rating Model Precision (%) Recall (%) F1(%) Accuracy (%)

SVM 89.53 97.05 93.14 92.73

(9)

Decision-Tree 89.61 99.98 93.61 93.29

Figure 1. Graphical representation Accuracy Evaluation Conclusion

In this project, a novel approval structure with unexpected security preservation and non- repudiation for FOG-VANET has been proposed, Noticing the earnestness of insane driving all over town nowadays, we proposed a defensive making system mindful of proactively recognize and tell the threats from advancing toward wild vehicles. To all the more probable fathom the consequence of absurd driving, we first theoretically surveyed the mishap probability of a typical stupid driving move, i.e., insane way developing. Considering the checked the necessity for defensive thoughtless driving cautions, we further developed an AI based Deep Neural Network (DNN)component by consolidating the figuring limits of neighbor vehicles what's more, a cloud laborer. By utilizing the noticed information from neighbor vehicles, the DNN instrument therefore additionally, cleverly recognizes insane vehicles. To help a particularly defensive disturbing structure, a three-level system configuration was made from existing vehicular associations. Furthermore, we devised a transmission load decline scheme by in a manner of speaking moving the data of driving moves with stupid potential. Considering the gathered noticing data, the cloud laborer all around the planet rates every vehicle's driving presentation by Support Vector Machine (SVM) and Decision Tree (DT) decision tree AI models We finally executed the proposed DNN segment into SUMO test framework. Expansive re-sanctioning outcomes addressed that our watched advised structure can decisively recognize stupid vehicles and give perfect alerts.

References

[1] Hoh, B. (2019). Safeguarding Privacy in Q. Luo, Y. Cao, J. Liu, and A. Bensilmane.

Limitation and route in self-sufficient driving: Threats and countermeasures. IEEE Communications Magazine.

[2] Wang, J., Zhao, L., Liu, J., & Kato, N. (2019). Savvy asset allotment for versatile edge processing: A profound fortification learning approach. IEEE Transactions on Emerging Topics in Computing.

[3] Zhao, L., Wang, J., Liu, J., & Kato, N. (2019). Directing for swarm the executives in shrewd urban areas: A profound support learning

(10)

point of view. IEEE Communications Magazine, 57(4), 88-93.

[4] Vennila, V., & Rajiv Kannan, A. (2019). Hybrid Parallel Linguistic Fuzzy Rules with Canopy MapReduce for Big Data Classification in Cloud. International Journal of Fuzzy Systems (IJFS), 21(3).

[5] Sikander, G., & Anwar, S. (2018). Driver weariness location frameworks: An audit. IEEE Transactions on Intelligent Transportation Systems.

[6] Li, B., Yan, J., Wu, W., Zhu, Z., & Hu, X. (2018). Elite visual following siamese locale proposition organization. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 8971-8980.

[7] Notices, A., Psannis, K.E., Ishibashi, Y., Kim, B.G., & Gupta, B.B. (2018). A proficient calculation for media-based observation framework (eamsus) in iot shrewd city structure.

Future Generation Computer Systems, 83, 619-628.

[8] Wang, J., Liu, J., & Kato, N. (2018). Systems administration and correspondences in independent driving: An overview. IEEE Communications Surveys and Tutorials, 21(2), 1243-1274.

[9] Gao, J., Zhao, L., & Shen, X. (2018). Organization utility boost dependent on an impetus instrument for honest announcing of neighborhood data. IEEE Transactions on Vehicular Technology, 67(8), 7523-7537.

[10] Matousek, M., Yassin, M., Van Der Heijden, R., Kargl, F. (2018). Strong discovery of irregular driving conduct. In IEEE 87th Vehicular Technology Conference (VTC Spring), 1-5.

[11] Balamurugan, P., Shyamala, D., & Sharmila, V. (2018). An energy minimizing score based optimal data gathering in wireless sensor networks. International Journal of Engineering & Technology, 7(2), 161-164.

[12] Balamurugan, P., Devi, M.S., & Sharmila, V. (2018). Detecting malicious nodes using data aggregation protocols in wireless sensor networks. International Journal of Engineering & Technology, 7(1.1), 594-598.

[13] Vennila, V., & Kannan, A.R. (2017). Discretized Support Vector Prediction Classifier for Big Data Computation and Information Sharing in Cloud. Asian Journal of Research in Social Sciences and Humanities, 7(2), 566-584.

[14] Martinez, C.M., Heucke, M., Wang, F.Y., Gao, B., & Cao D. (2017). Driving style acknowledgment for wise vehicle control and progressed driver help: An overview. IEEE Transactions on Intelligent Transportation Systems, 19(3), 666–676.

[15] Fadlullah, Z.M., Tang, F., Mao, B., Kato, N., Akashi, O., Inoue, T., & Mizutani, K.

(2017). Cutting edge profound getting the hang of: Evolving machine knowledge toward days to come shrewd organization traffic signal frameworks. IEEE Communications Surveys and Tutorials, 19(4), 2432–2455.

[16] Mahmud, S.S., Ferreira, L., Hoque, M.S., & Tavassoli, A. (2017). Use of proximal proxy pointers for security assessment: An audit of ongoing turns of events and examination

(11)

needs. IATSS research, 41(4), 153–163.

[17] Sharmila, V., Arasu, G.T., & Balamurugan, P. (2016). Non-Class Element based Iterative Text Clustering Algorithm for Improved Clustering Accuracy using Semantic Ontology. Asian Journal of Research in Social Sciences and Humanities, 6(cs1), 245-257.

[18] BalaMurugan, P., Ravichandran, T., & Sharmila, V. (2016). Grade and Energy based Data Gathering Protocols in Wireless Sensor Networks. Asian Journal of Research in Social Sciences and Humanities, 6(8), 728-744.

[19] Sharmila, V., Balamurugan, P., Vennila, V., & Savitha S. (2016). Information Retrieval and Recommendation Framework Using Maximum Matched Pattern Based Topic Models. International Journal of Innovative Research in Engineering Science and Technology.

[20] Vennila, V., & Kannan, A.R. (2016). Symmetric Matrix-based Predictive Classifier for Big Data computation and information sharing in Cloud. Computers & Electrical Engineering, 56, 831-841.

[21] Yu, J., Chen, Z., Zhu, Y., Chen, Y.J., Kong, L., & Li, M. (2016). Finegrained strange driving practices recognition and distinguishing proof with cell phones. IEEE exchanges on versatile figuring, 16(8), 2198–2212

[22] Luvizon, D.C., Nassu, B.T., & Minetto, R. (2016). A video-based framework for vehicle speed estimation in metropolitan streets. IEEE Transactions on Intelligent Transportation Systems, 18(6), 1393–1404.

[23] Kato, N., Fadlullah, Z.M., Mao, B., Tang, F., Akashi, O., Inoue, T., & Mizutani, K.

(2016). The profound learning vision for heterogeneous organization traffic light:

Proposal, difficulties, and future point of view. IEEE remote interchanges, 24(3), 146–

153.