Security Risks in Large Language Models and General Mitigation Strategies

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Since the inception of ChatGPT, Large Language Models (LLMs) have been reshaping many tasks such as document creation, decision support, computer coding and multi-language translations. Industries like education, healthcare and financial services are being transformed. However, their pervasiveness has also introduced security risks. In this study, a review of literature published in the most recent years is conducted to identify and categorise the potential risks associated with LLMs. Potential mitigation strategies and actionable solutions are discussed for each category. This study concludes by pointing out more efforts and future directions for the responsible development and use of LLMs.

Original languageEnglish
Title of host publicationAdvancing Innovative Cybersecurity Solutions and Approaches to Protect Digital Ecosystems - 1st IFIP TC 9, TC 11 International Conference in Cybersecurity, IFIP-UNIVEN-CSIR ICC 2025, Proceedings
EditorsJabu Mtsweni, Jackie Phahlamohlaka, Modimowabarwa Kanyane, Willard Munyoka, Kerry-Lynn Thomson, Lynn Futcher, Joey Jansen van Vuuren
PublisherSpringer Science and Business Media Deutschland GmbH
Pages3-12
Number of pages10
ISBN (Print)9783032130747
DOIs
Publication statusPublished - 2026
Externally publishedYes
Event1st IFIP TC 9, TC 11 International Conference in Cybersecurity, IFIP-UNIVEN-CSIR ICC 2025 - Tshwane, South Africa
Duration: 11 Dec 202512 Dec 2025

Publication series

NameIFIP Advances in Information and Communication Technology
Volume777 IFIPAICT
ISSN (Print)1868-4238
ISSN (Electronic)1868-422X

Conference

Conference1st IFIP TC 9, TC 11 International Conference in Cybersecurity, IFIP-UNIVEN-CSIR ICC 2025
Country/TerritorySouth Africa
CityTshwane
Period11/12/2512/12/25

Keywords

  • LLM security risks
  • Large Language Models
  • mitigation strategies

ASJC Scopus subject areas

  • Information Systems and Management

Fingerprint

Dive into the research topics of 'Security Risks in Large Language Models and General Mitigation Strategies'. Together they form a unique fingerprint.

Cite this