Overview on Deep Learning Hardware Attacks
| المصدر: | مجلة الدراسات المستدامة |
|---|---|
| الناشر: | الجمعية العلمية للدراسات التربوية المستدامة |
| المؤلف الرئيسي: | Ismaeel, Maath F. (Author) |
| مؤلفين آخرين: | Hammood, Maytham M. (Co-Author) , Alasad, Qutaiba (Co-Author) |
| المجلد/العدد: | مج5, ع4 |
| محكمة: | نعم |
| الدولة: |
العراق |
| التاريخ الميلادي: |
2023
|
| التاريخ الهجري: | 1445 |
| الشهر: | سبتمبر |
| الصفحات: | 183 - 195 |
| ISSN: |
2663-2284 |
| رقم MD: | 1410001 |
| نوع المحتوى: | بحوث ومقالات |
| اللغة: | الإنجليزية |
| قواعد المعلومات: | EduSearch |
| مواضيع: | |
| رابط المحتوى: |
عدد مرات التحميل
6
| المستخلص: |
Neural networks (NNs) are vital for diverse applications like speech/image recognition, cyber defense, decision-making, and financial forecasting. Deep neural networks (DNNs) are popular for their accuracy, but they demand high memory, power, and complexity. Hardware attacks (e.g., piracy, Trojans) threaten deep learning platforms. Securing hardware is crucial to safeguarding software integrity. This paper reviews the attacks on deep learning devices and accelerators, briefly reviewing their risks and consequences. Specific attacks (backdoor insertion, model extraction, spoofing, etc.) are addressed through secure hardware design, memory protection, code integrity checks, and secure boot processes. Hardware accelerators for DNNs enhance performance but face security risks. Ensuring confidentiality and integrity against potential threats is crucial. Hardware Trojans (HTs) and other attacks (fault injection, reverse engineering, sidechannel) are significant concerns. Effective strategies are necessary to mitigate these threats. Continued research and collaboration among hardware designers, AI/ML researchers, and cybersecurity experts are vital to build a secure foundation for global IC design flows. This enables safe deployment of AI/ML technologies in real-world scenarios. |
|---|---|
| ISSN: |
2663-2284 |
عناصر مشابهة
-
Hand Gesture Recognition System for Arabic Letters and Numbers Using Deep Learning
بواسطة: Saeed, Mogeeb A. منشور: (2021) -
Controlling a Robotic Arm Using Image-Based Feedback and Deep Reinforcement Learning
بواسطة: الزبط، عبدالله محمد منشور: (2019) -
Adopting Deep Learning Networks to Enhance Knowledge Management and Decision-Making in Business Industry
بواسطة: اللافي، خالد خلف محمد منشور: (2024) -
Deep Learning-Based Model for Arabic Fake News Detection Using Hybrid Contextual Features
بواسطة: تركي، حسين محمد منشور: (2023) -
La Prédiction du Diabète Utilisant le Deep Learning
بواسطة: Hima, Khaoula منشور: (2021)
