Detection of adversarial attacks against security systems based on deep learning model

Jaber, Mohanad J.; Jaber, Zahraa Jasim; Obaid, Ahmed J.

doi:10.47974/JDMSC-1935

Open Access Research Article

Detection of adversarial attacks against security systems based on deep learning model

Mohanad J. Jabermohanad.altalqani@uokufa.edu.iqDepartment of Basic ScienceCollege of DentistryKufa UniversityNajaf, IraqView full profile → , ^*Zahraa Jasim JaberCorresponding authorzahraaj.altalakany@uokufa.edu.iqDepartment of Computer ScienceFaculty of Computer Science and MathematicsKufa UniversityNajaf, IraqView full profile → , Ahmed J. Obaidahmedj.aljanaby@uokufa.edu.iqDepartment of Computer ScienceFaculty of Computer Science and MathematicsKufa UniversityNajaf, IraqView full profile →

* Corresponding author · click or hover a name for details

pp. 1523–1538Vol. 27Issue 5August 2024DOI: 10.47974/JDMSC-1935 Crossmark XML

Received:: 14 Nov 2023
Published Online:: 26 Aug 2024
Article type:: Research Article
Language:: EN
Article no.:: JDMSC-1935
Pages:: 1523–1538

Abstract

Many real-world problems can be solved using deep learning, including image processing and speech recognition. However, deep learning and neural networks are usually black boxes, making them vulnerable to threats from abnormal examples, such as adversarial and backdoor examples crafted by attackers with malicious intentions and, even worse, unintentionally mislabeled examples. The detection of such abnormal examples is therefore urgent and important. Various approaches have been used to detect specific types of abnormal examples, but all have limitations. Despite these limitations, this problem remains a significant one. To make AdvGAN more suitable for adversarial training, we propose updating its loss terms and testing it against various attacks. In contrast to generative models with no adversarial training, generative models with adversarial training are better at boosting adversarial robustness. Experimental results demonstrate a high true-positive rate (98.11%) and a low false-positive rate (3%).

Keywords

DDO attacks Security channel Attacks detection Adversarial attacks

Subject Classifications

68P2568P01

References

[1] S. P. Yadav, M. Jindal, P. Rani, V. H. C. De Albuquerque, C. Dos Santos Nascimento, and M. Kuma”An improved deep learning-based optimal object detection system from imag” Multimed. Tools Appl., Sep. (2023), doi: 10.1007/s11042-023-16736-5.
[2] C. Szegedy et al “Intriguing properties of neural network” ArXiv Prepr. ArXiv13126199, (2013).
[3] P. Rani and R. Sharm “Intelligent transportation system for internet of vehicles based vehicular networks for smart citi” Comput. Electr. Eng., vol. 105, p. 108543 (2023).
[4] N. Hussain, P. Rani, H. Chouhan, and U. S. Gau”Cyber Security and Privacy of Connected and Automated Vehicles (CAVs)-Based Federated Learning: Challenges, Opportunities, and Open Issu” in Federated Learning for IoT Applications, S. P. Yadav, B. S. Bhati, D. P. Mahato, and S. Kumar, Eds., in EAI/Springer Innovations in Communication and Computing., Cham: Springer International Publishing (2022), pp. 169–183. doi: 10.1007/978-3-030-85559-8_11.
[5] A. Kurakin, I. J. Goodfellow, and S. Bengi “Adversarial examples in the physical wor” in Artificial intelligence safety and security, Chapman and Hall/CRC, pp. 99–112 (2018).
[6] M. Sharif, S. Bhagavatula, L. Bauer, and M. K. Reite”Accessorize to a Crime: Real and Stealthy Attacks on State-of-the-Art Face Recogniti” in Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security, Vienna Austria: ACM, pp. 1528–1540, Oct. (2016). doi: 10.1145/2976749.2978392.
[7] Z. Che et al “Adversarial attack against deep saliency models powered by non-redundant prio” IEEE Trans. Image Process., vol. 30, pp. 1973–1988 (2021).
[8] P. Rani, P. N. Singh, S. Verma, N. Ali, P. K. Shukla, and M. Alhassa”An implementation of modified blowfish technique with honey bee behavior optimization for load balancing in cloud system environme” Wirel. Commun. Mob. Comput., vol. 2022, pp. 1–14 (2022).
[9] P. Rani et al “Federated Learning-Based Misbehaviour Detection for the 5G-Enabled Internet of Vehicl” IEEE Trans. Consum. Electron., pp. 1–1 (2023), doi: 10.1109/TCE.2023.3328020.
[10] I. J. Goodfellow, J. Shlens, and C. Szeged “Explaining and harnessing adversarial exampl” ArXiv Prepr. ArXiv14126572 (2014).
[11] N. Papernot, P. McDaniel, X. Wu, S. Jha, and A. Swam “Distillation as a defense to adversarial perturbations against deep neural networ” in 2016 IEEE symposium on security and privacy (SP), IEEE, pp. 582–597 (2016).
[12] N. Papernot, P. McDaniel, S. Jha, M. Fredrikson, Z. B. Celik, and A. Swam “The limitations of deep learning in adversarial settin” in 2016 IEEE European symposium on security and privacy (EuroS&P), IEEE, pp. 372–387 (2016).
[13] N. Carlini and D. Wagne “r” “Towards evaluating the robustness of neural networ” in 2017 ieee symposium on security and privacy (sp), Ieee, pp. 39–57 (2017).
[14] J. H. Metzen, T. Genewein, V. Fischer, and B. Bischof “On detecting adversarial perturbatio” ArXiv Prepr. ArXiv170204267 (2017).
[15] R. Herrera-Pereda, A. Taboada Crispi, D. Babin, W. Philips, and M. Holsbach Cost”A Review On digital image processing techniques for in-Vivo confocal images of the corn” Med. Image Anal., vol. 73, p. 102188, Oct. (2021), doi: 10.1016/j.media.2021.102188.
[16] S. R. Low “Pattern Recognition in Handwriti” in Series on Language Processing, Pattern Recognition, and Intelligent Systems, vol. 06, WORLD SCIENTIFIC, pp. 77–95 (2021). doi: 10.1142/9789811239014_0005.
[17] S. Tabik, R. F. Alvear-Sandoval, M. M. Ruiz, J.-L. Sancho-Gómez, A. R. Figueiras-Vidal, and F. Herrer “MNIST-NET10: A heterogeneous deep networks fusion based on the degree of certainty to reach 0.1% error rate. Ensembles overview and propos” Inf. Fusion, vol. 62, pp. 73–80, Oct. (2020), doi: 10.1016/j.inffus.2020.04.002.
[18] S. Wan, S. Ding, and C. Che “Edge computing enabled video segmentation for real-time traffic monitoring in internet of vehicl” Pattern Recognit., vol. 121, p. 108146, Jan. (2022), doi: 10.1016/j.patcog.2021.108146.
[19] Y. Wang et al “Improvement of MNIST image recognition based on C” in IOP Conference Series: Earth and Environmental Science, IOP Publishing, p. 012097 (2020).
[20] S. R. Kulkarni and B. Rajendra “Spiking neural networks for handwritten digit recognition—Supervised learning and network optimizati” Neural Netw., vol. 103, pp. 118–127, Jul. (2018), doi: 10.1016/j.neunet.2018.03.019.
[21] V. Nguyen, J. Cai, and J. Ch “Hybrid CNN-GRU model for high efficient handwritten digit recogniti” in Proceedings of the 2nd International Conference on Artificial Intelligence and Pattern Recognition, Beijing China: ACM, pp. 66–71 Aug. (2019). doi: 10.1145/3357254.3357276.
[22] Z. Jan and B. Verm “Multiple strong and balanced cluster-based ensemble of deep learne” Pattern Recognit., vol. 107, p. 107420, Nov. (2020), doi: 10.1016/j.patcog.2020.107420.
[23] W. Xia, X. Zhang, Q. Gao, and X. Ga “Adversarial self-supervised clustering with cluster-specificity distributi” Neurocomputing, vol. 449, pp. 38–47, Aug. (2021), doi: 10.1016/j.neucom.2021.03.108.
[24] W. Liang, Y. Wu, M. Li, and Y. Ca “Adaptive multiple kernel fusion model using spatial-statistical information for high resolution SAR image classificati” Neurocomputing, vol. 492, pp. 382–395, Jul. (2022), doi: 10.1016/j.neucom.2022.03.062.
[25] S. S. Kadam, A. C. Adamuthe, and A. B. Pati “CNN model for image classification on MNIST and fashion-MNIST datas” J. Sci. Res., vol. 64, no. 2, pp. 374–384 (2020).
[26] Z. Xue, M. Zhang, Y. Liu, and P. D “Attention-based second-order pooling network for hyperspectral image classificati” IEEE Trans. Geosci. Remote Sens., vol. 59, no. 11, pp. 9600–9615 (2021).
[27] J. Chen, X. Lin, Z. Shi, and Y. Liu “Link prediction adversarial attack via iterative gradient atta” IEEE Trans. Comput. Soc. Syst., vol. 7, no. 4, pp. 1081–1094 (2020).
[28] J. Li, T. Xie, L. Chen, F. Xie, X. He, and Z. Zhen “Adversarial attack on large scale gra”” IEEE Trans. Knowl. Data Eng., vol. 35, no. 1, pp. 82–95 (2021).
[29] H. Wang, G. Li, X. Liu, and L. Li “A Hamiltonian Monte Carlo Method for Probabilistic Adversarial Attack and Learni” IEEE Trans. Pattern Anal. Mach. Intell., vol. 44, no. 4, pp. 1725–1737, Apr. (2022), doi: 10.1109/TPAMI.2020.3032061.
[30] L. Schwinn et al “Dynamically sampled nonlocal gradients for stronger adversarial attac” in 2021 International Joint Conference on Neural Networks (IJCNN), IEEE, pp. 1–8 (2021).
[31] S. Chen, Z. He, C. Sun, J. Yang, and X. Huan “Universal adversarial attack on attention and the resulting dataset damagen” IEEE Trans. Pattern Anal. Mach. Intell., vol. 44, no. 4, pp. 2188–2197 (2020).
[32] Y. Yu, X. Gao, and C.-Z. X “Lafeat: Piercing through adversarial defenses with latent featur” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 5735–5745 (2021).
[33] Y. Xiang, Y. Xu, Y. Li, W. Ma, Q. Xuan, and Y. Li “Side-Channel Gray-Box Attack for DN” IEEE Trans. Circuits Syst. II Express Briefs, vol. 68, no. 1, pp. 501–505, Jan. (2021), doi: 10.1109/TCSII.2020.3012005.
[34] H. Wang, S. Wang, Z. Jin, Y. Wang, C. Chen, and M. Tistarell “Similarity-based Gray-box Adversarial Attack Against Deep Face Recogniti” in 2021 16th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2021), pp. 1–8 Dec. (2021). doi: 10.1109/FG52635.2021.9667076.
[35] Z. Liu, Y. Luo, L. Wu, S. Li, Z. Liu, and S. Z. Li “Are Gradients on Graph Structure Reliable in Gray-box Attack” in Proceedings of the 31st ACM International Conference on Information & Knowledge Management, in CI’22. New York, NY, USA: Association for Computing Machinery, pp. 1360–1368 Oct. (2022). doi: 10.1145/3511808.3557238.
[36] A. Fawzi, O. Fawzi, and P. Frossar “Analysis oclassifiers’’ robustness to adversarial perturbations. Co” ArXiv Prepr. ArXiv150202590, (2015).
[37] S. Shen, G. Jin, K. Gao, and Y. Zhan “Ape-gan: Adversarial perturbation elimination with g” ArXiv Prepr. ArXiv170705474 (2017).
[38] F. Tramèr, A. Kurakin, N. Papernot, I. Goodfellow, D. Boneh, and P. McDani”l “Ensemble adversarial training: Attacks and defens” ArXiv Prepr. ArXiv170507204 (2017).
[39] W. Xu, D. Evans, and Y. Q “Feature squeezing: Detecting adversarial examples in deep neural networ” ArXiv Prepr. ArXiv170401155 (2017).
[40] D. Meng and H. Che “Magnet: a two-pronged defense against adversarial exampl” in Proceedings of the 2017 ACM SIGSAC conference on computer and communications security, pp. 135–147 (2017).
[41] I. Goodfellow et al. “Generative adversarial ne” Adv. Neural Inf. Process. Syst., vol. 27 (2014).
[42] Y. Liu, X. Chen, C. Liu, and D. Son “Delving into transferable adversarial examples and black-box attac” ArXiv Prepr. ArXiv161102770, (2016).
[43] Jawad, Zainab A. & Obaid, Ahmed J. A systemic literature overview of Fake News Challenge (FNC-1) dataset and its use in fake news detection schemes, Journal of Discrete Mathematical Sciences and Cryptography, 1-10, DOI: 10.47974/JDMSC-1567.
[44] Abdulbaqi, Azmi Shawkat , Muslim, Israa Falih , Al-Ameer, Asraa A. Abd & Obaid, Ahmed J. () Healthcare surveillance based on cloud computing utilizing mobile devices, Journal of Discrete Mathematical Sciences and Cryptography, 1-8, DOI: 10.47974/JDMSC-1566.

Views: 285Downloads: 155Citations: 0