MEF: A Capability-Aware Multi-Encryption Framework for Evaluating Vulnerabilities in Black-Box Large Language Models Article Swipe

PDF

Mingyu Yu , Sheng Wang , Yu Wei , Su‐Juan Qin , Fei Gao , Wenmin Li ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2505.23404

Recent advancements in adversarial jailbreak attacks have exposed critical vulnerabilities in Large Language Models (LLMs), enabling the circumvention of alignment safeguards through increasingly sophisticated prompt manipulations. Based on our experiments, we found that the effectiveness of jailbreak strategies is influenced by the comprehension ability of the attacked LLM. Building on this insight, we propose a capability-aware Multi-Encryption Framework (MEF) for evaluating vulnerabilities in black-box LLMs. Specifically, MEF first categorizes the comprehension ability level of the LLM, then applies different strategies accordingly: For models with limited comprehension ability, MEF adopts the Fu+En1 strategy, which integrates layered semantic mutations with an encryption technique, more effectively contributing to evasion of the LLM's defenses at the input and inference stages. For models with strong comprehension ability, MEF uses a more complex Fu+En1+En2 strategy, in which additional dual-ended encryption techniques are applied to the LLM's responses, further contributing to evasion of the LLM's defenses at the output stage. Experimental results demonstrate the effectiveness of our approach, achieving attack success rates of 98.9% on GPT-4o (29 May 2025 release) and 99.8% on GPT-4.1 (8 July 2025 release). Our work contributes to a deeper understanding of the vulnerabilities in current LLM alignment mechanisms.

Related Topics

National Institutional Ranking Framework

The Myth Of The Framework

Cynefin Framework

Django (Web Framework)

Iran Nuclear Deal Framework

United Nations Framework Convention On Climate Change

Bootstrap (Front-End Framework)

Electron (Software Framework)

Flask (Web Framework)

Metal–Organic Framework

The Open Group Architecture Framework

Robot Framework

Concepts

No concepts available.

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2505.23404
PDF: https://arxiv.org/pdf/2505.23404
OA Status: green
OpenAlex ID: https://openalex.org/W4414876843

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4414876843

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2505.23404

Digital Object Identifier
Title: MEF: A Capability-Aware Multi-Encryption Framework for Evaluating Vulnerabilities in Black-Box Large Language Models

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-05-29

Full publication date if available
Authors: Mingyu Yu, Sheng Wang, Yu Wei, Su‐Juan Qin, Fei Gao, Wenmin Li

List of authors in order
Landing page: https://arxiv.org/abs/2505.23404

Publisher landing page
PDF URL: https://arxiv.org/pdf/2505.23404

Direct link to full text PDF
Open access: Yes

Whether a free full text is available
OA status: green

Open access status per OpenAlex
OA URL: https://arxiv.org/pdf/2505.23404

Direct OA link when available
Cited by: 0

Total citation count in OpenAlex

Full payload

id	https://openalex.org/W4414876843
doi	https://doi.org/10.48550/arxiv.2505.23404
ids.doi	https://doi.org/10.48550/arxiv.2505.23404
ids.openalex	https://openalex.org/W4414876843
fwci
type	preprint
title	MEF: A Capability-Aware Multi-Encryption Framework for Evaluating Vulnerabilities in Black-Box Large Language Models
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
topics[0].id	https://openalex.org/T11614
topics[0].field.id	https://openalex.org/fields/17
topics[0].field.display_name	Computer Science
topics[0].score	0.7516000270843506
topics[0].domain.id	https://openalex.org/domains/3
topics[0].domain.display_name	Physical Sciences
topics[0].subfield.id	https://openalex.org/subfields/1710
topics[0].subfield.display_name	Information Systems
topics[0].display_name	Cloud Data Security Solutions
topics[1].id	https://openalex.org/T10764
topics[1].field.id	https://openalex.org/fields/17
topics[1].field.display_name	Computer Science
topics[1].score	0.7028999924659729
topics[1].domain.id	https://openalex.org/domains/3
topics[1].domain.display_name	Physical Sciences
topics[1].subfield.id	https://openalex.org/subfields/1702
topics[1].subfield.display_name	Artificial Intelligence
topics[1].display_name	Privacy-Preserving Technologies in Data
topics[2].id	https://openalex.org/T11424
topics[2].field.id	https://openalex.org/fields/17
topics[2].field.display_name	Computer Science
topics[2].score	0.6579999923706055
topics[2].domain.id	https://openalex.org/domains/3
topics[2].domain.display_name	Physical Sciences
topics[2].subfield.id	https://openalex.org/subfields/1702
topics[2].subfield.display_name	Artificial Intelligence
topics[2].display_name	Security and Verification in Computing
is_xpac	False
apc_list
apc_paid
language	en
locations[0].id	pmh:oai:arXiv.org:2505.23404
locations[0].is_oa	True
locations[0].source.id	https://openalex.org/S4306400194
locations[0].source.issn
locations[0].source.type	repository
locations[0].source.is_oa	True
locations[0].source.issn_l
locations[0].source.is_core	False
locations[0].source.is_in_doaj	False
locations[0].source.display_name	arXiv (Cornell University)
locations[0].source.host_organization	https://openalex.org/I205783295
locations[0].source.host_organization_name	Cornell University
locations[0].source.host_organization_lineage	https://openalex.org/I205783295
locations[0].license
locations[0].pdf_url	https://arxiv.org/pdf/2505.23404
locations[0].version	submittedVersion
locations[0].raw_type	text
locations[0].license_id
locations[0].is_accepted	False
locations[0].is_published	False
locations[0].raw_source_name
locations[0].landing_page_url	http://arxiv.org/abs/2505.23404
locations[1].id	doi:10.48550/arxiv.2505.23404
locations[1].is_oa	True
locations[1].source.id	https://openalex.org/S4306400194
locations[1].source.issn
locations[1].source.type	repository
locations[1].source.is_oa	True
locations[1].source.issn_l
locations[1].source.is_core	False
locations[1].source.is_in_doaj	False
locations[1].source.display_name	arXiv (Cornell University)
locations[1].source.host_organization	https://openalex.org/I205783295
locations[1].source.host_organization_name	Cornell University
locations[1].source.host_organization_lineage	https://openalex.org/I205783295
locations[1].license
locations[1].pdf_url
locations[1].version
locations[1].raw_type	article
locations[1].license_id
locations[1].is_accepted	False
locations[1].is_published
locations[1].raw_source_name
locations[1].landing_page_url	https://doi.org/10.48550/arxiv.2505.23404
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5083118115
authorships[0].author.orcid	https://orcid.org/0000-0001-6730-7478
authorships[0].author.display_name	Mingyu Yu
authorships[0].author_position	first
authorships[0].raw_author_name	Yu, Mingyu
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5100371313
authorships[1].author.orcid	https://orcid.org/0000-0003-2696-2171
authorships[1].author.display_name	Sheng Wang
authorships[1].author_position	middle
authorships[1].raw_author_name	Wang, Wei
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5008125230
authorships[2].author.orcid	https://orcid.org/0000-0001-8473-6856
authorships[2].author.display_name	Yu Wei
authorships[2].author_position	middle
authorships[2].raw_author_name	Wei, Yanjie
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5040230786
authorships[3].author.orcid	https://orcid.org/0000-0002-6405-6711
authorships[3].author.display_name	Su‐Juan Qin
authorships[3].author_position	middle
authorships[3].raw_author_name	Qin, Sujuan
authorships[3].is_corresponding	False
authorships[4].author.id	https://openalex.org/A5100318655
authorships[4].author.orcid	https://orcid.org/0000-0001-6739-5076
authorships[4].author.display_name	Fei Gao
authorships[4].author_position	middle
authorships[4].raw_author_name	Gao, Fei
authorships[4].is_corresponding	False
authorships[5].author.id	https://openalex.org/A5108781441
authorships[5].author.orcid
authorships[5].author.display_name	Wenmin Li
authorships[5].author_position	last
authorships[5].raw_author_name	Li, Wenmin
authorships[5].is_corresponding	False
has_content.pdf	True
has_content.grobid_xml	True
is_paratext	False
open_access.is_oa	True
open_access.oa_url	https://arxiv.org/pdf/2505.23404
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	MEF: A Capability-Aware Multi-Encryption Framework for Evaluating Vulnerabilities in Black-Box Large Language Models
has_fulltext	True
is_retracted	False
updated_date	2025-11-06T06:51:31.235846
primary_topic.id	https://openalex.org/T11614
primary_topic.field.id	https://openalex.org/fields/17
primary_topic.field.display_name	Computer Science
primary_topic.score	0.7516000270843506
primary_topic.domain.id	https://openalex.org/domains/3
primary_topic.domain.display_name	Physical Sciences
primary_topic.subfield.id	https://openalex.org/subfields/1710
primary_topic.subfield.display_name	Information Systems
primary_topic.display_name	Cloud Data Security Solutions
cited_by_count	0
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2505.23404
best_oa_location.is_oa	True
best_oa_location.source.id	https://openalex.org/S4306400194
best_oa_location.source.issn
best_oa_location.source.type	repository
best_oa_location.source.is_oa	True
best_oa_location.source.issn_l
best_oa_location.source.is_core	False
best_oa_location.source.is_in_doaj	False
best_oa_location.source.display_name	arXiv (Cornell University)
best_oa_location.source.host_organization	https://openalex.org/I205783295
best_oa_location.source.host_organization_name	Cornell University
best_oa_location.source.host_organization_lineage	https://openalex.org/I205783295
best_oa_location.license
best_oa_location.pdf_url	https://arxiv.org/pdf/2505.23404
best_oa_location.version	submittedVersion
best_oa_location.raw_type	text
best_oa_location.license_id
best_oa_location.is_accepted	False
best_oa_location.is_published	False
best_oa_location.raw_source_name
best_oa_location.landing_page_url	http://arxiv.org/abs/2505.23404
primary_location.id	pmh:oai:arXiv.org:2505.23404
primary_location.is_oa	True
primary_location.source.id	https://openalex.org/S4306400194
primary_location.source.issn
primary_location.source.type	repository
primary_location.source.is_oa	True
primary_location.source.issn_l
primary_location.source.is_core	False
primary_location.source.is_in_doaj	False
primary_location.source.display_name	arXiv (Cornell University)
primary_location.source.host_organization	https://openalex.org/I205783295
primary_location.source.host_organization_name	Cornell University
primary_location.source.host_organization_lineage	https://openalex.org/I205783295
primary_location.license
primary_location.pdf_url	https://arxiv.org/pdf/2505.23404
primary_location.version	submittedVersion
primary_location.raw_type	text
primary_location.license_id
primary_location.is_accepted	False
primary_location.is_published	False
primary_location.raw_source_name
primary_location.landing_page_url	http://arxiv.org/abs/2505.23404
publication_date	2025-05-29
publication_year	2025
referenced_works_count	0
abstract_inverted_index.a	54, 124, 185
abstract_inverted_index.(8	177
abstract_inverted_index.an	98
abstract_inverted_index.at	110, 149
abstract_inverted_index.by	40
abstract_inverted_index.in	2, 10, 62, 129, 191
abstract_inverted_index.is	38
abstract_inverted_index.of	18, 35, 44, 73, 106, 145, 158, 165, 188
abstract_inverted_index.on	27, 49, 167, 175
abstract_inverted_index.to	104, 137, 143, 184
abstract_inverted_index.we	30, 52
abstract_inverted_index.(29	169
abstract_inverted_index.For	81, 116
abstract_inverted_index.LLM	193
abstract_inverted_index.MEF	66, 87, 122
abstract_inverted_index.May	170
abstract_inverted_index.Our	181
abstract_inverted_index.and	113, 173
abstract_inverted_index.are	135
abstract_inverted_index.for	59
abstract_inverted_index.our	28, 159
abstract_inverted_index.the	16, 33, 41, 45, 69, 74, 89, 107, 111, 138, 146, 150, 156, 189
abstract_inverted_index.2025	171, 179
abstract_inverted_index.July	178
abstract_inverted_index.LLM,	75
abstract_inverted_index.LLM.	47
abstract_inverted_index.have	6
abstract_inverted_index.more	101, 125
abstract_inverted_index.that	32
abstract_inverted_index.then	76
abstract_inverted_index.this	50
abstract_inverted_index.uses	123
abstract_inverted_index.with	83, 97, 118
abstract_inverted_index.work	182
abstract_inverted_index.(MEF)	58
abstract_inverted_index.98.9%	166
abstract_inverted_index.99.8%	174
abstract_inverted_index.Based	26
abstract_inverted_index.LLM's	108, 139, 147
abstract_inverted_index.LLMs.	64
abstract_inverted_index.Large	11
abstract_inverted_index.first	67
abstract_inverted_index.found	31
abstract_inverted_index.input	112
abstract_inverted_index.level	72
abstract_inverted_index.rates	164
abstract_inverted_index.which	92, 130
abstract_inverted_index.Fu+En1	90
abstract_inverted_index.GPT-4o	168
abstract_inverted_index.Models	13
abstract_inverted_index.Recent	0
abstract_inverted_index.adopts	88
abstract_inverted_index.attack	162
abstract_inverted_index.deeper	186
abstract_inverted_index.models	82, 117
abstract_inverted_index.output	151
abstract_inverted_index.prompt	24
abstract_inverted_index.stage.	152
abstract_inverted_index.strong	119
abstract_inverted_index.(LLMs),	14
abstract_inverted_index.GPT-4.1	176
abstract_inverted_index.ability	43, 71
abstract_inverted_index.applied	136
abstract_inverted_index.applies	77
abstract_inverted_index.attacks	5
abstract_inverted_index.complex	126
abstract_inverted_index.current	192
abstract_inverted_index.evasion	105, 144
abstract_inverted_index.exposed	7
abstract_inverted_index.further	141
abstract_inverted_index.layered	94
abstract_inverted_index.limited	84
abstract_inverted_index.propose	53
abstract_inverted_index.results	154
abstract_inverted_index.stages.	115
abstract_inverted_index.success	163
abstract_inverted_index.through	21
abstract_inverted_index.Building	48
abstract_inverted_index.Language	12
abstract_inverted_index.ability,	86, 121
abstract_inverted_index.attacked	46
abstract_inverted_index.critical	8
abstract_inverted_index.defenses	109, 148
abstract_inverted_index.enabling	15
abstract_inverted_index.insight,	51
abstract_inverted_index.release)	172
abstract_inverted_index.semantic	95
abstract_inverted_index.Framework	57
abstract_inverted_index.achieving	161
abstract_inverted_index.alignment	19, 194
abstract_inverted_index.approach,	160
abstract_inverted_index.black-box	63
abstract_inverted_index.different	78
abstract_inverted_index.inference	114
abstract_inverted_index.jailbreak	4, 36
abstract_inverted_index.mutations	96
abstract_inverted_index.release).	180
abstract_inverted_index.strategy,	91, 128
abstract_inverted_index.Fu+En1+En2	127
abstract_inverted_index.additional	131
abstract_inverted_index.dual-ended	132
abstract_inverted_index.encryption	99, 133
abstract_inverted_index.evaluating	60
abstract_inverted_index.influenced	39
abstract_inverted_index.integrates	93
abstract_inverted_index.responses,	140
abstract_inverted_index.safeguards	20
abstract_inverted_index.strategies	37, 79
abstract_inverted_index.technique,	100
abstract_inverted_index.techniques	134
abstract_inverted_index.adversarial	3
abstract_inverted_index.categorizes	68
abstract_inverted_index.contributes	183
abstract_inverted_index.demonstrate	155
abstract_inverted_index.effectively	102
abstract_inverted_index.mechanisms.	195
abstract_inverted_index.Experimental	153
abstract_inverted_index.accordingly:	80
abstract_inverted_index.advancements	1
abstract_inverted_index.contributing	103, 142
abstract_inverted_index.experiments,	29
abstract_inverted_index.increasingly	22
abstract_inverted_index.Specifically,	65
abstract_inverted_index.circumvention	17
abstract_inverted_index.comprehension	42, 70, 85, 120
abstract_inverted_index.effectiveness	34, 157
abstract_inverted_index.sophisticated	23
abstract_inverted_index.understanding	187
abstract_inverted_index.manipulations.	25
abstract_inverted_index.vulnerabilities	9, 61, 190
abstract_inverted_index.Multi-Encryption	56
abstract_inverted_index.capability-aware	55
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	6
citation_normalized_percentile