paper-with-me

Papers

Can Small Language Models Reliably Resist Jailbreak Attacks? A Comprehensive Evaluation

2025-03-09 · Wenhui Zhang, Huiyu Xu, Zhibo Wang, Zeqing He, Ziqi Zhu, Kui Ren

Small language models (SLMs) have emerged as promising alternatives to large language models (LLMs) due to their low computational demands, enhanced privacy guarantees and comparable performance in specific domains through light-weight fine-tuning. Deploying SLMs on edge devices, such as smartphones and smart vehicles, has become a growing trend. However, the security implications of SLMs have received less attention than LLMs, particularly regarding jailbreak attacks, which is recognized as one of the top threats of LLMs by the OWASP. In this paper, we conduct the first large-scale empirical study of SLMs' vulnerabilities to jailbreak attacks. Through systematically evaluation on 63 SLMs from 15 mainstream SLM families against 8 state-of-the-art jailbreak methods, we demonstrate that 47.6% of evaluated SLMs show high susceptibility to jailbreak attacks (ASR > 40%) and 38.1% of them can not even resist direct harmful query (ASR > 50%). We further analyze the reasons behind the vulnerabilities and identify four key factors: model size, model architecture, training datasets and training techniques. Moreover, we assess the effectiveness of three prompt-level defense methods and find that none of them achieve perfect performance, with detection accuracy varying across different SLMs and attack methods. Notably, we point out that the inherent security awareness play a critical role in SLM security, and models with strong security awareness could timely terminate unsafe response with little reminder. Building upon the findings, we highlight the urgent need for security-by-design approaches in SLM development and provide valuable insights for building more trustworthy SLM ecosystem.

📄 PDF Abstract BibTeX arXiv:2503.06519

Code (0)

등록된 구현이 없습니다.

Methods 이 논문이 사용한 방법론

Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…
Attention 설명 없음

Similar Papers 제목 키워드 기반

SmoothLLM: Defending Large Language Models Against Jailbreaking Attacks

2023-10-05 · Alexander Robey, Eric Wong, Hamed Hassani, George J. Pappas

Despite efforts to align large language models (LLMs) with human intentions, widely-used LLMs such as GPT, Llama, and Claude are susceptible to jailbreaking attacks, wherein an adversary fools a targeted LLM into generat…

Pruning for Protection: Increasing Jailbreak Resistance in Aligned LLMs Without Fine-Tuning

2024-01-19 · Adib Hasan, Ileana Rugina, Alex Wang

This paper investigates the impact of model compression on the way Large Language Models (LLMs) process prompts, particularly concerning jailbreak resistance. We show that moderate WANDA pruning can enhance resistance to…

Model Compression

PrisonBreak: Jailbreaking Large Language Models with Fewer Than Twenty-Five Targeted Bit-flips

2024-12-10 · Zachary Coalson, Jeonghyun Woo, Yu Sun, Shiyang Chen 외

We introduce a new class of attacks on commercial-scale (human-aligned) language models that induce jailbreaking through targeted bitwise corruptions in model parameters. Our adversary can jailbreak billion-parameter lan…

Computational Efficiency

Voice Jailbreak Attacks Against GPT-4o

2024-05-29 · Xinyue Shen, Yixin Wu, Michael Backes, Yang Zhang

Recently, the concept of artificial assistants has evolved from science fiction into real-world applications. GPT-4o, the newest multimodal large language model (MLLM) across audio, vision, and text, has further blurred …

Language ModellingLarge Language ModelMultimodal Large Language Model

Multi-Turn Jailbreaks Are Simpler Than They Seem

2025-08-11 · Xiaoxue Yang, Jaeha Lee, Anna-Katharina Dick, Jasper Timm 외 arxiv

While defenses against single-turn jailbreak attacks on Large Language Models (LLMs) have improved significantly, multi-turn jailbreaks remain a persistent vulnerability, often achieving success rates exceeding 70% again…