TY - GEN
T1 - The Hidden Dangers of Publicly Accessible LLMs
T2 - 15th EAI International Conference on Digital Forensics and Cyber Crime, ICDF2C 2024
AU - Vaishnav, Lakshika
AU - Singh, Sakshi
AU - Cornell, Kimberly A.
N1 - Publisher Copyright: © ICST Institute for Computer Sciences, Social Informatics and Telecommunications Engineering 2025.
PY - 2025
Y1 - 2025
N2 - In recent years, the use of large language models (LLMs) for information retrieval has become as ubiquitous as using search engines like Google. However, the widespread adoption of these sophisticated AI models, such as Gab AI, introduces significant risks due to their open-source nature and massive scale. Gab AI promotes itself as an unbiased platform, yet it provides an opportunity for users to exploit the LLM for malicious purposes. This paper explores the literature surrounding the malevolent use of LLMs and investigates how open-source platforms like Gab AI can be manipulated to generate harmful content, orchestrate attack plans, and more. By examining the potential misuse of readily accessible LLMs like Gab AI, which, unlike many nefarious tools, do not require access via the dark web, this study aims to foster awareness and prompt discussions on mitigating the risks associated with these powerful technologies.
AB - In recent years, the use of large language models (LLMs) for information retrieval has become as ubiquitous as using search engines like Google. However, the widespread adoption of these sophisticated AI models, such as Gab AI, introduces significant risks due to their open-source nature and massive scale. Gab AI promotes itself as an unbiased platform, yet it provides an opportunity for users to exploit the LLM for malicious purposes. This paper explores the literature surrounding the malevolent use of LLMs and investigates how open-source platforms like Gab AI can be manipulated to generate harmful content, orchestrate attack plans, and more. By examining the potential misuse of readily accessible LLMs like Gab AI, which, unlike many nefarious tools, do not require access via the dark web, this study aims to foster awareness and prompt discussions on mitigating the risks associated with these powerful technologies.
KW - Attack Prompts
KW - Cybersecurity
KW - Malicious LLMs
UR - https://www.scopus.com/pages/publications/105006882086
U2 - 10.1007/978-3-031-89363-6_18
DO - 10.1007/978-3-031-89363-6_18
M3 - Conference contribution
SN - 9783031893629
T3 - Lecture Notes of the Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, LNICST
SP - 312
EP - 330
BT - Digital Forensics and Cyber Crime - 15th EAI International Conference, ICDF2C 2024, Proceedings
A2 - Goel, Sanjay
A2 - Uzun, Ersin
A2 - Xie, Mengjun
A2 - Sarkar, Sumantra
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 9 October 2024 through 10 October 2024
ER -