AdversaFlow Visual Red Teaming for Large Language Models with Multi-Level Adversarial Flow

AdversaFlow : Visual Red Teaming for Large Language Models with Multi-Level Adversarial Flow

Large Language Models (LLMs) are powerful but also raise significant security concerns, particularly regarding the harm they can cause, such as generating fake news that manipulates public opinion on social media and providing responses to unethical activities. Traditional red teaming approaches for...

Ausführliche Beschreibung

Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on visualization and computer graphics. - 1996. - PP(2024) vom: 16. Sept.
1. Verfasser: Deng, Dazhen (VerfasserIn)
Weitere Verfasser: Zhang, Chuhan, Zheng, Huawei, Pu, Yuwen, Ji, Shouling, Wu, Yingcai
Format: Online-Aufsatz
Sprache:English
Veröffentlicht: 2024
Zugriff auf das übergeordnete Werk:IEEE transactions on visualization and computer graphics
Schlagworte:Journal Article
LEADER 01000caa a22002652c 4500
001 NLM377708208
003 DE-627
005 20250306161059.0
007 cr uuu---uuuuu
008 240917s2024 xx |||||o 00| ||eng c
024 7 |a 10.1109/TVCG.2024.3456150  |2 doi 
028 5 2 |a pubmed25n1258.xml 
035 |a (DE-627)NLM377708208 
035 |a (NLM)39283796 
040 |a DE-627  |b ger  |c DE-627  |e rakwb 
041 |a eng 
100 1 |a Deng, Dazhen  |e verfasserin  |4 aut 
245 1 0 |a AdversaFlow  |b Visual Red Teaming for Large Language Models with Multi-Level Adversarial Flow 
264 1 |c 2024 
336 |a Text  |b txt  |2 rdacontent 
337 |a ƒaComputermedien  |b c  |2 rdamedia 
338 |a ƒa Online-Ressource  |b cr  |2 rdacarrier 
500 |a Date Revised 16.09.2024 
500 |a published: Print-Electronic 
500 |a Citation Status Publisher 
520 |a Large Language Models (LLMs) are powerful but also raise significant security concerns, particularly regarding the harm they can cause, such as generating fake news that manipulates public opinion on social media and providing responses to unethical activities. Traditional red teaming approaches for identifying AI vulnerabilities rely on manual prompt construction and expertise. This paper introduces AdversaFlow, a novel visual analytics system designed to enhance LLM security against adversarial attacks through human-AI collaboration. AdversaFlow involves adversarial training between a target model and a red model, featuring unique multi-level adversarial flow and fluctuation path visualizations. These features provide insights into adversarial dynamics and LLM robustness, enabling experts to identify and mitigate vulnerabilities effectively. We present quantitative evaluations and case studies validating our system's utility and offering insights for future AI security solutions. Our method can enhance LLM security, supporting downstream scenarios like social media regulation by enabling more effective detection, monitoring, and mitigation of harmful content and behaviors 
650 4 |a Journal Article 
700 1 |a Zhang, Chuhan  |e verfasserin  |4 aut 
700 1 |a Zheng, Huawei  |e verfasserin  |4 aut 
700 1 |a Pu, Yuwen  |e verfasserin  |4 aut 
700 1 |a Ji, Shouling  |e verfasserin  |4 aut 
700 1 |a Wu, Yingcai  |e verfasserin  |4 aut 
773 0 8 |i Enthalten in  |t IEEE transactions on visualization and computer graphics  |d 1996  |g PP(2024) vom: 16. Sept.  |w (DE-627)NLM098269445  |x 1941-0506  |7 nnas 
773 1 8 |g volume:PP  |g year:2024  |g day:16  |g month:09 
856 4 0 |u http://dx.doi.org/10.1109/TVCG.2024.3456150  |3 Volltext 
912 |a GBV_USEFLAG_A 
912 |a SYSFLAG_A 
912 |a GBV_NLM 
912 |a GBV_ILN_350 
951 |a AR 
952 |d PP  |j 2024  |b 16  |c 09