Evolving Jailbreaks: Automated Multi-Objective Long-Tail Attacks on Large Language Models
Identification
- Public canonical key
- radar:r-daa1fd02ba5f3aadd17373b25508aaa2
- Authors
- Wenjing Hong · Zhonghua Rong · Li Wang · Feng Chang · Jian Zhu · Ke Tang · Zexuan Zhu · Yew-Soon Ong
- Reported year
- 2026
- Reported language
- en
- Source type
- article
- First detected
- 2026-11-14T02:45:52.452452+00:00
- Last metadata observation
- 2026-11-14T02:45:52.452452+00:00
- Last HTTP verification
- Not checked
Institutions reported by the source
Shenzhen University · University of Science and Technology of China · Southern University of Science and Technology · Agency for Science, Technology and Research
Reported affiliation countries
CN · SG
Documentary abstract — original language
Large Language Models (LLMs) have been widely deployed, especially through free Web-based applications that expose them to diverse user-generated inputs, including those from long-tail distributions such as low-resource languages and encrypted private data. This open-ended exposure increases the risk of jailbreak attacks that undermine model safety alignment. While recent studies have shown that leveraging long-tail distributions can facilitate such jailbreaks, existing approaches largely rely on handcrafted rules, limiting the systematic evaluation of these security and privacy vulnerabilities. In this work, we present EvoJail, an automated framework for discovering long-tail distribution attacks via multi-objective evolutionary search. EvoJail formulates long-tail attack prompt generation as a multi-objective optimization problem that jointly maximizes attack effectiveness and minimizes output perplexity, and introduces a semantic-algorithmic solution representation to capture both high-level semantic intent and low-level structural transformations of encryption-decryption logic. Building upon this representation, EvoJail integrates LLM-assisted operators into a multi-objective evolutionary framework, enabling adaptive and semantically informed mutation and crossover for efficiently exploring a highly structured and open-ended search space. Extensive experiments demonstrate that EvoJail consistently discovers diverse and effective long-tail jailbreak strategies, achieving competitive performance with existing methods in both individual and ensemble level.
Identifiers
- OPENALEX
- https://openalex.org/W7140140067
- DOI
- 10.1109/tevc.2026.3726531
Documentary context
- Documentary family
- article
- Source-reported types
- article
- Publishers
- Institute of Electrical and Electronics Engineers
Entity mentions are not evidence of legal recognition. Jurisdiction studied is separate from author affiliation country.
Sources
OpenAlex · Record 1
Cite this reference
Wenjing Hong; Zhonghua Rong; Li Wang; Feng Chang; Jian Zhu; Ke Tang; Zexuan Zhu; Yew-Soon Ong. Evolving Jailbreaks: Automated Multi-Objective Long-Tail Attacks on Large Language Models. 2026. https://doi.org/10.1109/tevc.2026.3726531
JSON · CSL-JSON · BibTeX · RIS · CSV · JSON-LD ·
APA, Chicago, ISO 690: bibliographic formats using available fields. OSCOLA / Bluebook: structured data only; full legal citation compliance is not claimed without verification of required elements.
- APA
- Wenjing Hong; Zhonghua Rong; Li Wang; Feng Chang; Jian Zhu; Ke Tang; Zexuan Zhu; Yew-Soon Ong. (2026). Evolving Jailbreaks: Automated Multi-Objective Long-Tail Attacks on Large Language Models. https://doi.org/10.1109/tevc.2026.3726531
- Chicago
- Wenjing Hong; Zhonghua Rong; Li Wang; Feng Chang; Jian Zhu; Ke Tang; Zexuan Zhu; Yew-Soon Ong. Evolving Jailbreaks: Automated Multi-Objective Long-Tail Attacks on Large Language Models. 2026. https://doi.org/10.1109/tevc.2026.3726531
- ISO 690
- Wenjing Hong; Zhonghua Rong; Li Wang; Feng Chang; Jian Zhu; Ke Tang; Zexuan Zhu; Yew-Soon Ong. Evolving Jailbreaks: Automated Multi-Objective Long-Tail Attacks on Large Language Models. 2026. https://doi.org/10.1109/tevc.2026.3726531
Warning: Undefined array key "last_link_verified" in /home/clients/e27d81e12f04d2bfd747aa264540ac56/sites/_UDTR_APP/legal_library_reference.php on line 31
UDTR Global Rights of Nature Legal Library. Evolving Jailbreaks: Automated Multi-Objective Long-Tail Attacks on Large Language Models. Documentary record: https://www.declarationuniverselledesdroitsdelarbre.org/global-rights-of-nature-radar/reference/r-daa1fd02ba5f3aadd17373b25508aaa2/ · Last HTTP verification: not performed
Research dossiers and documentary relations
Source roles
Bibliographic / original publication source · https://doi.org/10.1109/tevc.2026.3726531
An authority link alone does not verify a legal claim. Abstracts and full texts remain subject to their source terms.
Documentary status
Relevance pending
Library
This identity is not yet included in the Legal Library.