Enhancing The Detection Of Adversarial Attacks Using Deep Learning Neural Transformer Models
Open Access DepositedIn today’s world, cybersecurity and artificial intelligence (AI) are crucial to ensuring the confidentiality, integrity, and availability of interconnected digital systems for human safety and productivity. The rapid adoption of AI in mission-critical systems has led to the development of highly sophisticated cyberattacks with malicious intent, including sabotage, theft, and potential threats to human life (Payne et al., 2024).This Doctor of Engineering praxis presents a novel study and produces an applied machine learning model that utilizes advancements in AI, such as Bidirectional Encoder Representations from Transformers (BERT), Neural Machine Translation (NMT), and Extreme Gradient Boosting (XGBoost), to counter adversarial machine learning (AML) attacks. The praxis examines techniques used for AML, identifies strategies to mitigate attacks, and evaluates the potential of using Transformer Models for effective prevention and detection. It focuses on identifying attack techniques such as evasion, noise, poisoning, malicious code injection, and detection, aiming to take explicit methods to nullify these attacks. Additionally, by using an ensemble methodology that integrates Large Language Models (LLMs), deep learning (DL), and boosted decision trees, the study presents a unique approach to enhancing the security of AI defense systems. Organizations with safety and mission-critical systems can adopt this approach to build additional defense against adversarial attacks.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
Gunasekara_gwu_0075A_17086.pdf | 2025-04-11 | Open Access |
|