Saved in:
Bibliographic Details
Main Authors: Gaure, Simen, Koffas, Stefanos, Picek, Stjepan, Rønjom, Sondre
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.15652
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909338043416576
author Gaure, Simen
Koffas, Stefanos
Picek, Stjepan
Rønjom, Sondre
author_facet Gaure, Simen
Koffas, Stefanos
Picek, Stjepan
Rønjom, Sondre
contents Large Language Models (LLMs) have gained significant popularity recently. LLMs are susceptible to various attacks but can also improve the security of diverse systems. However, besides enabling more secure systems, how well do open source LLMs behave as covertext distributions to, e.g., facilitate censorship-resistant communication? In this paper, we explore open-source LLM-based covert channels. We empirically measure the security vs. capacity of an open-source LLM model (Llama-7B) to assess its performance as a covert channel. Although our results indicate that such channels are not likely to achieve high practical bitrates, we also show that the chance for an adversary to detect covert communication is low. To ensure our results can be used with the least effort as a general reference, we employ a conceptually simple and concise scheme and only assume public models.
format Preprint
id arxiv_https___arxiv_org_abs_2405_15652
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle $$\mathbf{L^2\cdot M = C^2}$$ Large Language Models are Covert Channels
Gaure, Simen
Koffas, Stefanos
Picek, Stjepan
Rønjom, Sondre
Cryptography and Security
Large Language Models (LLMs) have gained significant popularity recently. LLMs are susceptible to various attacks but can also improve the security of diverse systems. However, besides enabling more secure systems, how well do open source LLMs behave as covertext distributions to, e.g., facilitate censorship-resistant communication? In this paper, we explore open-source LLM-based covert channels. We empirically measure the security vs. capacity of an open-source LLM model (Llama-7B) to assess its performance as a covert channel. Although our results indicate that such channels are not likely to achieve high practical bitrates, we also show that the chance for an adversary to detect covert communication is low. To ensure our results can be used with the least effort as a general reference, we employ a conceptually simple and concise scheme and only assume public models.
title $$\mathbf{L^2\cdot M = C^2}$$ Large Language Models are Covert Channels
topic Cryptography and Security
url https://arxiv.org/abs/2405.15652