Citation :

AnNing, Mazida Ahmad, Huda lbrahim, 2024. "Adversarial Examples in Deep Learning: Understanding and Mitigating Vulnerabilities" ESP International Journal of Advancements in Science & Technology (ESP-IJAST)  Volume 2, Issue 1: 22-26.

Abstract :

Adversarial examples have become a critical concern in deep learning systems due to their ability to deceive models with imperceptible perturbations. This paper focuses on understanding and mitigating vulnerabilities caused by adversarial examples. To achieve this, we first investigate the background and reasons behind the existence of adversarial examples. Then, we propose and implement different defence methods, including adversarial training and defensive distillation. These methods are evaluated on various benchmark datasets, and the results demonstrate their effectiveness in improving robustness against adversarial attacks. Furthermore, we analyze the limitations and potential further research directions in this field. Overall, this study contributes to a better understanding of the characteristics, impacts, and countermeasures of adversarial examples in deep learning systems.

References :

[1] DT Ha.Line outage vulnerabilities of power systems: models and indicators[D].,2018

[2] CES Agustin.Mitigating Deep Learning Vulnerabilities from Adversarial Examples Attack in the Cybersecurity Domain[D].,2019

[3] X Yuan, P He, Q Zhu, et al.Adversarial Examples: Attacks and Defenses for Deep Learning[D].IEEE Transactions on Neural Networks & Learning Systems,2019

[4] K Wang,F Li,CM Chen,et al.Interpreting Adversarial Examples and Robustness for Deep Learning-Based Auto-Driving Systems[D].IEEE Transactions on Intelligent Transportation Systems,2022

[5] AM Algarni.Quantitative economics of security: Software vulnerabilities and data breaches.[D].,2016

[6] DY Meng.Generating deep learning adversarial examples in black-box scenario[D].Electronic Design Engineering,2018

[7] OH, Ahmad.A Systems Approach to Understanding and Mitigating Barriers to Travel Accessibility and Well-being in the West Bank, Palestine.[D].,2015

[8] P Sermanet.A Deep Learning Pipeline for Image Understanding and Acoustic Modeling.[D].,2014

[9] J Wang,C Wang,Q Lin,et al.Adversarial attacks and defenses in deep learning for image recognition: A survey[D].Neurocomputing,2022

[10] B Deng,Z Ran,J Chen,et al.Adversarial Examples Generation Algorithm through DCGAN[D].Intelligent Automation & Soft Computing,2021

[11] R Haffar,N Jebreel,D Sanchez,et al.Generating Deep Learning Model-Specific Explanations at the End User's Side[D].,2022

[12] Kolja Stahl,Andrea Graziadei,Therese Dau,et al.Protein structure prediction with in-cell photo-crosslinking mass spectrometry and deep learning[D].,2023

[13] S Kokalj-Filipovic,R Miller.Adversarial Examples in RF Deep Learning: Detection of the Attack and its Physical Robustness[D].,2019

[14] PK Douglas, F Vasheghani Farahani.On the Similarity of Deep Learning Representations Across Didactic and Adversarial Examples[D].,2020

[15] S Hussain, P Neekhara, S Dubnov, et al.WaveGuard: Understanding and Mitigating Audio Adversarial Examples[D].,2021

Keywords :

Adversarial Examples, Deep Learning, Vulnerabilities, Defence Methods, Robustness.