English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

With the advancement of Large Language Models (LLMs), significant progresshas been made in code generation, enabling LLMs to transform natural languageinto programming code. These Code LLMs have been widely accepted by massiveusers and organizations. However, a dangerous nature is hidden in the code,which is the existence of fatal vulnerabilities. While some LLM providers haveattempted to address these issues by aligning with human guidance, theseefforts fall short of making Code LLMs practical and robust. Without a deepunderstanding of the performance of the LLMs under the practical worst cases,it would be concerning to apply them to various real-world applications. Inthis paper, we answer the critical issue: Are existing Code LLMs immune togenerating vulnerable code? If not, what is the possible maximum severity ofthis issue in practical deployment scenarios? In this paper, we introduceDeceptPrompt, a novel algorithm that can generate adversarial natural languageinstructions that drive the Code LLMs to generate functionality correct codewith vulnerabilities. DeceptPrompt is achieved through a systematicevolution-based algorithm with a fine grain loss design. The unique advantageof DeceptPrompt enables us to find natural prefix/suffix with totally benignand non-directional semantic meaning, meanwhile, having great power in inducingthe Code LLMs to generate vulnerable code. This feature can enable us toconduct the almost-worstcase red-teaming on these LLMs in a real scenario,where users are using natural language. Our extensive experiments and analyseson DeceptPrompt not only validate the effectiveness of our approach but alsoshed light on the huge weakness of LLMs in the code generation task. Whenapplying the optimized prefix/suffix, the attack success rate (ASR) willimprove by average 50% compared with no prefix/suffix applying.

DeceptPrompt: Exploiting LLM-driven Code Generation via Adversarial Natural Language Instructions