STUDY AI VULNERABILITIES SUCH AS ADVERSARIAL ATTACKS AND DATA POISONING IN EDUCATION TESTING SYSTEM.

Qobulov D. B. & Yuldashev N. A.


Annotatsiya

The integration of artificial intelligence (AI) technologies into high-level educational assessments in higher education poses significant cybersecurity vulnerabilities that threaten academic integrity, student privacy, and institutional reputation. This paper provides a comprehensive analysis of the cybersecurity challenges faced by AI-based automated online testing systems in higher education. By systematically examining the application of machine learning techniques, this study identifies significant vulnerabilities, such as evasion attacks, model manipulation, data poisoning, and label modification attacks, that can undermine the reliability and fairness of AI-based assessments. The study uses a mixed-methods approach that combines theoretical analysis with simulation-based experiments to assess the susceptibility of components of a common AI testing system to various attack vectors.


Adabiyotlar

  1. Chen, L., & Xie, P. (2021). Artificial intelligence in education: A review. IEEE Access, 9, 75264-75278.
  2. Foltýnek, T., Meuschke, N., & Gipp, B. (2020). Academic plagiarism detection: A systematic literature review. ACM Computing Surveys, 52(6), 1-42.
  3. Garcia, R., Falkner, K., & Vitale, J. (2023). AI in educational assessment: Opportunities and challenges. Computers and Education: Artificial Intelligence, 4, 100123.
  4. Milano, S., McGrane, J. A., & Leonelli, M. (2021). AI and responsible innovation in education. AI & Society, 36(3), 913-925.
  5. Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7), 100779.
  6. N. Yuldashev (2025). The Development Of Automated Creating Test Questions Using Artificial Intelligence: Knowledge Estimation” International Journal of Pedagogics, ISSN 2771-2281, Pages: 185-189, Volume 05 Issue 06 June 2025.