{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,8]],"date-time":"2026-05-08T00:18:26Z","timestamp":1778199506055,"version":"3.51.4"},"reference-count":0,"publisher":"Association for the Advancement of Artificial Intelligence (AAAI)","issue":"5","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["AAAI"],"abstract":"<jats:p>Large Language Models (LLMs) are increasingly vulnerable to adversarial prompts that exploit semantic ambiguities\nto bypass safety mechanisms, resulting in harmful or inappropriate outputs. Such attacks, including jailbreaking and\nprompt injection, pose significant risks to the integrity and\navailability of LLMs in security-critical applications. This paper proposes the Adversarial Prompt Disentanglement (APD)\nframework, a novel defense mechanism that proactively identifies and neutralizes malicious components in input prompts\nbefore they are processed by the LLM. The APD framework\nintegrates three key innovations: (1) a mutual information-\nbased semantic decomposition method to isolate adversarial and benign prompt components, ensuring statistical in-\ndependence; (2) a graph-based intent classification approach\nthat leverages spectral analysis to detect malicious patterns\nin prompt semantics; and (3) a lightweight transformer-based\nclassifier trained on real-world datasets of toxic and jailbreaking prompts, enabling efficient and accurate adversarial intent detection. Evaluated on diverse datasets containing adversarial prompts, APD demonstrates superior robustness, reducing harmful output generation by over 85% while maintaining negligible impact on model performance. The framework\u2019s computational efficiency supports real-time deploy-\nment, making it a practical solution for securing LLMs. Our\nwork addresses critical challenges in machine learning security on novel attacks and integrity methods for ML systems,\nand offers a scalable, ethically grounded defense against\nprompt-based adversarial threats.<\/jats:p>","DOI":"10.1609\/aaai.v40i5.37389","type":"journal-article","created":{"date-parts":[[2026,3,17]],"date-time":"2026-03-17T23:08:03Z","timestamp":1773788883000},"page":"3876-3884","source":"Crossref","is-referenced-by-count":1,"title":["Disentangling Adversarial Prompts: A Semantic-Graph Defense for Robust LLM Security"],"prefix":"10.1609","volume":"40","author":[{"given":"Xiang","family":"Fang","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Wanlong","family":"Fang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"9382","published-online":{"date-parts":[[2026,3,14]]},"container-title":["Proceedings of the AAAI Conference on Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/download\/37389\/41351","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/download\/37389\/41351","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,17]],"date-time":"2026-03-17T23:08:03Z","timestamp":1773788883000},"score":1,"resource":{"primary":{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/view\/37389"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,14]]},"references-count":0,"journal-issue":{"issue":"5","published-online":{"date-parts":[[2026,3,17]]}},"URL":"https:\/\/doi.org\/10.1609\/aaai.v40i5.37389","relation":{},"ISSN":["2374-3468","2159-5399"],"issn-type":[{"value":"2374-3468","type":"electronic"},{"value":"2159-5399","type":"print"}],"subject":[],"published":{"date-parts":[[2026,3,14]]}}}