Adversarial Machine Learning: A Review of Methods, Tools, and Critical Industry Sectors

Strategic Insights -

Researchers from the National Technical University of Athens have discussed the rapid advancement of Artificial Intelligence (AI), particularly Machine Learning (ML) and Deep Learning (DL), which has led to the development of high-performance models in various applications. However, security threats arise from the vulnerabilities of ML models to adversarial attacks and data poisoning, posing risks such as system malfunctions and decision errors. This study has been funded by the HORIZON EUROPE Framework Programme and the Smart Networks and Services Joint Undertaking (SNS JU) under the European Union's Horizon Europe research and innovation programme.

Key Takeaways:

  • The researchers have conducted a comprehensive survey of the Adversarial Machine Learning (AML) landscape in modern AI systems, focusing on the dual aspects of robustness and privacy.
  • The study explores adversarial attacks and defenses using taxonomies and investigates robustness benchmarks alongside open-source AML technologies and software tools.
  • The researchers have analyzed attacks, defenses, and evaluation concepts in four industry fields: automotive, digital healthcare, electrical power and energy systems (EPES), and Large Language Model (LLM)-based Natural Language Processing (NLP) systems.
  • The study has identified specific risks such as system malfunctions and decision errors due to vulnerabilities in ML models to adversarial attacks and data poisoning.
  • The researchers have emphasized the importance of robustness and privacy preservation in the future, promoting a holistic view of the modern AI-reliant industry.
  • The study has been peer-reviewed and published in the Artificial Intelligence Review journal.
  • Sotiris Pelekis, Thanos Koutroubas, Afroditi Blika, Evangelos Karakolis, Christos Ntanos, Evangelos Spiliotis, Dimitris Askounis, and Anastasis Berdelis are among the additional authors of this research.

A total of 4 industry fields were analyzed in this study.

The concept of Adversarial Machine Learning (AML) was explored in the context of modern AI systems, with a focus on robustness and privacy preservation.

The study identified specific risks in the following industry sectors:

+ Automotive

+ Digital healthcare

+ Electrical power and energy systems (EPES)

+ Large Language Model (LLM)-based Natural Language Processing (NLP) systems

The researchers have identified 4 industry fields as being heavily reliant on AI systems:

+ Automotive

+ Digital healthcare

+ Electrical power and energy systems (EPES)

+ Large Language Model (LLM)-based Natural Language Processing (NLP) systems

A total of 2 funding agencies supported this research:

+ HORIZON EUROPE Framework Programme

+ Smart Networks and Services Joint Undertaking (SNS JU)

Statistics:

  • The study was funded by 2 agencies.
  • A total of 4 industry fields were analyzed in this study.
  • The researchers identified specific risks in the following industry sectors: automotive, digital healthcare, electrical power and energy systems (EPES), and Large Language Model (LLM)-based Natural Language Processing (NLP) systems.
  • A total of 8 authors contributed to this research, including the lead author Sotiris Pelekis.

Sources:

  • NewsRx. Recent Findings from National Technical University of Athens Has Provided New Information about Machine Learning (Adversarial Machine Learning: a Review of Methods, Tools, and Critical Industry Sectors). Medical Patent Law Weekly. May 28, 2025; p 1985.
  • Adversarial Machine Learning: a Review of Methods, Tools, and Critical Industry Sectors. Artificial Intelligence Review, 2025; 58(8).