CNN and MLP neural network ensembles for packet classification and adversary defense

doi:10.23919/ICN.2020.0023

2021, Vol. 2

Issue (1): 66-82 doi: 10.23919/ICN.2020.0023

Networks

CNN and MLP neural network ensembles for packet classification and adversary defense

Bruce Hartpence^*(

),Andres Kwasinski(

)

GCCIS i-School at the Rochester Institute of Technology, Rochester, New York, NY 14623, USA
Department of Computer Engineering, KGCOE at the Rochester Institute of Technology, Rochester, New York, NY 14623, USA

Download:

PDF (12614 KB)

HTML
Export: BibTeX | EndNote (RIS)

Abstract

Machine learning techniques such as artificial neural networks are seeing increased use in the examination of communication network research questions. Central to many of these research questions is the need to classify packets and improve visibility. Multi-Layer Perceptron (MLP) neural networks and Convolutional Neural Networks (CNNs) have been used to successfully identify individual packets. However, some datasets create instability in neural network models. Machine learning can also be subject to data injection and misclassification problems. In addition, when attempting to address complex communication network challenges, extremely high classification accuracy is required. Neural network ensembles can work towards minimizing or even eliminating some of these problems by comparing results from multiple models. After ensembles tuning, training time can be reduced, and a viable and effective architecture can be obtained. Because of their effectiveness, ensembles can be utilized to defend against data poisoning attacks attempting to create classification errors. In this work, ensemble tuning and several voting strategies are explored that consistently result in classification accuracy above 99%. In addition, ensembles are shown to be effective against these types of attack by maintaining accuracy above 98%.

Key words： Convolutional Neural Network (CNN) Multi-Layer Perception (MLP) ensemble classification adversary

Received: 03 September 2020 Online: 19 August 2021

Corresponding Authors: Bruce Hartpence E-mail: bhhics@rit.edu;axkeec@rit.edu

About author: Bruce Hartpence received the PhD degree in computing and information sciences from the Rochester Institute of Technology in 2020, and the MS degree in information technology from the same institute from in 1998. He is currently a professor for the GCCIS i-School at the Rochester Institute of Technology (RIT) where he teaches networking and communication. For the last several years, he has explored the application of neural networks to communication challenges. He has also authored several books and video series for O’Reilly publishing including the Packet Guide series. He is also a co-developer of the IEEE 1910.1 Meshed Tree standard for the IEEE. His current areas of research include neural network ensembles applied wired and wireless, real time communication, and intelligent networking.|Andres Kwasinski received the diploma degree in electrical engineering from the Buenos Aires Institute of Technology, Buenos Aires, Argentina, in 1992 and the MS and PhD degrees in electrical and computer engineering from the University of Maryland, College Park, MD, USA, in 2000 and 2004, respectively. He is currently a professor at the Department of Computer Engineering, Rochester Institute of Technology, Rochester, NY, USA. He has co-authored over 90 publications in peer-reviewed journals and international conferences. He has also co-authored the books Cooperative Communications and Networking (Cambridge University Press, 2009) and 3D Visual Communications (Wiley, 2013). His current areas of research include cognitive radios and wireless networks, cross-layer techniques in wireless communications, and signal processing applied to smart infrastructures. He is also the chief editor of the Signal Processing Repository (SigPort) and an area editor for Special Initiatives of the IEEE Signal Processing Magazine.


	Service

	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	Bruce Hartpence
	Andres Kwasinski

Cite this article:

Bruce Hartpence,Andres Kwasinski. CNN and MLP neural network ensembles for packet classification and adversary defense. , 2021, 2: 66-82.

URL:

http://icn.tsinghuajournals.com/10.23919/ICN.2020.0023 OR http://icn.tsinghuajournals.com/Y2021/V2/I1/66

Fig. 1 Architecture overview.

Table 1 Breakdown of the class structure.

Fig. 2 Raw packet vs. Cleaned packets.

Fig. 3 Packet image.

Fig. 4 Base structure of convolutional neural network.

Fig. 5 MLP neural network.

Fig. 6 General stage.

Fig. 7 UDP stage.

Table 2 MLP model configuration.

Table 3 CNN model configuration.

Table 4 Early model comparison.

Table 5 Elimination model comparison.

Table 6 Elimination model—1000 iterations.

Table 7 Modified models—1000 iterations.

Table 8 Adding a model—500 iterations.

Table 9 Final ensemble.

Table 10 Final UDP ensemble.

Table 11 Final TCP ensemble.

Table 12 Overall performance.

Table 13 Initial validation based weights.

Table 14 Final validation based weights.

Table 15 Validation weight performance.

Table 16 Initial model average based weights.

Table 17 Final model average based weights.

Table 18 Ensemble average weight performance.

Table 19 Actual model Ave Wt performance.

Table 20 Dataset 5—Layer 2.

Table 21 Dataset 5—Layers 3 and 4.

Table 22 Dataset 5—Poisoned.


[1]	Hartpence B. and Kwasinski A., Fast internet packet and flow classification based on artificial neural networks, in Proc. 2019 SoutheastCon, Huntsville, AL, USA, 2019, p. 19433953.
[2]	Hartpence B. and Kwasinski A., A convolutional neural network approach to improving network visibility, in Proc. 2020 29th Wireless and Optical Communications Conf. (WOCC), Newark, NJ, USA, 2020, pp. 1-6.
[3]	Sharkey A. J. C., Sharkey N. E., Gerecke U., and Chandroth G. O., The “test and select” approach to ensemble combination, in Proc. 1st Int. Workshop on Multiple Classifier Systems, Cagliari, Italy, 2000, pp. 30-44.
[4]	Dietterich T. G., Ensemble methods in machine learning, in Proc. 1st Int. Workshop on Multiple Classifier Systems, Cagliari, Italy, 2000, pp. 1-15.
[5]	Yaman M. A., Subasi A., and Rattay F., Comparison of random subspace and voting ensemble machine learning methods for face recognition, Symmetry, vol. 10, no. 11, p. 651, 2018.
[6]	Knauer U., Von Rekowski C. S., Stecklina M., Krokotsch T., Minh T. P., Hauffe V., Kilias D., Ehrhardt I., Sagischewski H., Chmara S., et al., Tree species classification based on hybrid ensembles of a convolutional neural network (CNN) and random forest classifiers, Remote Sens., vol. 11, no. 23, p. 2788, 2019.
[7]	Tasci E., Voting combinations-based ensemble of fine-tuned convolutional neural networks for food image recognition, Multimed. Tools Appl., vol. 79, no. 41, pp. 30 397-30 418, 2020.
[8]	Livieris I. E., Kanavos A., Tampakas V., and Pintelas P., A weighted voting ensemble self-labeled algorithm for the detection of lung abnormalities from x-rays, Algorithms, vol. 12, no. 3, p. 64, 2019.
[9]	Strauss T., Hanselmann M., Junginger A., and Ulmer H., Ensemble methods as a defense to adversarial perturbations against deep neural networks, arXiv preprint arXiv: 1709.03423, 2017.
[10]	Wei W. Q., Liu L., Loper M., Chow K. H., Gursoy E., Truex S., and Wu Y. Z., Cross-layer strategic ensemble defense against adversarial examples, in Proc. 2020 Int. Conf. Computing, Networking and Communications (ICNC), Big Island, HI, USA, 2020, pp. 456-460.
[11]	He K. M., Zhang X. Y., Ren S. Q., and Sun J., Deep residual learning for image recognition, in Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016, pp. 770-778.
[12]	Simonyan K. and Zisserman A., Very deep convolutional networks for large-scale image recognition, arXiv preprint arXiv: 1409.1556, 2014.
[13]	Pan B., Shi Z. W., and Xu X., Hierarchical guidance filtering-based ensemble classification for hyperspectral images, IEEE Trans. Geosci. Remote Sens., vol. 55, no. 7, pp. 4177-4189, 2017.
[14]	Gunderson S. and Jagodzinski F., Ensemble voting schemes that improve machine learning models for predicting the effects of protein mutations, in Proc. 2018 ACM Int. Conf. Bioinformatics, Computational Biology, and Health Informatics, New York, NY, USA, 2018, pp. 211-219.
[15]	Sin E. and Wang L. P., Bitcoin price prediction using ensembles of neural networks, in Proc. 2017 13th Int. Conf. Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNCFSKD), Guilin, China, 2017, pp. 666-671.
[16]	Xu L. Y., Zhou X., Ren Y. M., and Qin Y. F., A traffic classification method based on packet transport layer payload by ensemble learning, in Proc. 2019 IEEE Sympo. Computers and Communications (ISCC), Barcelona, Spain, 2019, pp. 1-6.
[17]	Gómez S. E., Hernández-Callejo L., Martínez B. C., and Sánchez-Esguevillas A. J., Exploratory study on class imbalance and solutions for network traffic classification, Neurocomputing, vol. 343, pp. 100-119, 2019.
[18]	Gómez S. E., Martínez B. C., Sánchez-Esguevillas A. J., and Callejo L. H., Ensemble network traffic classification: Algorithm comparison and novel ensemble scheme proposal, Comput. Networks, vol. 127, pp. 68-80, 2017.
[19]	Gargiulo F., Kuncheva L. I., and Sansone C., Network protocol verification by a classifier selection ensemble, in Proc. 8th Int. Workshop on Multiple Classifier Systems, Reykjavik, Iceland, 2009, pp. 314-323.
[20]	CAIDA, .
[21]	Szegedy C., Vanhoucke V., Ioffe S., Shlens J., and Wojna Z., Rethinking the inception architecture for computer vision, in Proc. 2016 IEEE Conf. Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016, pp. 2818-2826.
[22]	Tramèr F., Kurakin A., Papernot N., Goodfellow I., Boneh D., and McDaniel P., Ensemble adversarial training: Attacks and defenses, arXiv preprint arXiv: 1705.07204, 2017.
[23]	Nasr M., Bahramali A., and Houmansadr A., Blind adversarial network perturbations, arXiv preprint arXiv: 2002.06495, 2020.

No related articles found!

Viewed

Full text

Abstract

Cited

Shared

Discussed