Can AI-Generated Text be Reliably Detected?

Bibliographic Details
Title:	Can AI-Generated Text be Reliably Detected?
Authors:	Sadasivan, Vinu Sankar, Kumar, Aounon, Balasubramanian, Sriram, Wang, Wenxiao, Feizi, Soheil
Publication Year:	2023
Collection:	Computer Science
Subject Terms:	Computer Science - Computation and Language, Computer Science - Artificial Intelligence, Computer Science - Machine Learning
More Details:	Large Language Models (LLMs) perform impressively well in various applications. However, the potential for misuse of these models in activities such as plagiarism, generating fake news, and spamming has raised concern about their responsible use. Consequently, the reliable detection of AI-generated text has become a critical area of research. AI text detectors have shown to be effective under their specific settings. In this paper, we stress-test the robustness of these AI text detectors in the presence of an attacker. We introduce recursive paraphrasing attack to stress test a wide range of detection schemes, including the ones using the watermarking as well as neural network-based detectors, zero shot classifiers, and retrieval-based detectors. Our experiments conducted on passages, each approximately 300 tokens long, reveal the varying sensitivities of these detectors to our attacks. Our findings indicate that while our recursive paraphrasing method can significantly reduce detection rates, it only slightly degrades text quality in many cases, highlighting potential vulnerabilities in current detection systems in the presence of an attacker. Additionally, we investigate the susceptibility of watermarked LLMs to spoofing attacks aimed at misclassifying human-written text as AI-generated. We demonstrate that an attacker can infer hidden AI text signatures without white-box access to the detection method, potentially leading to reputational risks for LLM developers. Finally, we provide a theoretical framework connecting the AUROC of the best possible detector to the Total Variation distance between human and AI text distributions. This analysis offers insights into the fundamental challenges of reliable detection as language models continue to advance. Our code is publicly available at https://github.com/vinusankars/Reliability-of-AI-text-detectors. Comment: Published in Transactions on Machine Learning Research (TMLR)
Document Type:	Working Paper
Access URL:	http://arxiv.org/abs/2303.11156
Accession Number:	edsarx.2303.11156
Database:	arXiv

FullText	Text: Availability: 0 CustomLinks: – Url: http://arxiv.org/abs/2303.11156 Name: EDS - Arxiv Category: fullText Text: View this record from Arxiv MouseOverText: View this record from Arxiv – Url: https://resolver.ebsco.com/c/xy5jbn/result?sid=EBSCO:edsarx&genre=article&issn=&ISBN=&volume=&issue=&date=20230317&spage=&pages=&title=Can AI-Generated Text be Reliably Detected?&atitle=Can%20AI-Generated%20Text%20be%20Reliably%20Detected%3F&aulast=Sadasivan%2C%20Vinu%20Sankar&id=DOI: Name: Full Text Finder (for New FTF UI) (s8985755) Category: fullText Text: Find It @ SCU Libraries MouseOverText: Find It @ SCU Libraries
Header	DbId: edsarx DbLabel: arXiv An: edsarx.2303.11156 RelevancyScore: 1051 AccessLevel: 3 PubType: Report PubTypeId: report PreciseRelevancyScore: 1051.00512695313
IllustrationInfo
Items	– Name: Title Label: Title Group: Ti Data: Can AI-Generated Text be Reliably Detected? – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Sadasivan%2C+Vinu+Sankar%22">Sadasivan, Vinu Sankar</searchLink><br /><searchLink fieldCode="AR" term="%22Kumar%2C+Aounon%22">Kumar, Aounon</searchLink><br /><searchLink fieldCode="AR" term="%22Balasubramanian%2C+Sriram%22">Balasubramanian, Sriram</searchLink><br /><searchLink fieldCode="AR" term="%22Wang%2C+Wenxiao%22">Wang, Wenxiao</searchLink><br /><searchLink fieldCode="AR" term="%22Feizi%2C+Soheil%22">Feizi, Soheil</searchLink> – Name: DatePubCY Label: Publication Year Group: Date Data: 2023 – Name: Subset Label: Collection Group: HoldingsInfo Data: Computer Science – Name: Subject Label: Subject Terms Group: Su Data: <searchLink fieldCode="DE" term="%22Computer+Science+-+Computation+and+Language%22">Computer Science - Computation and Language</searchLink><br /><searchLink fieldCode="DE" term="%22Computer+Science+-+Artificial+Intelligence%22">Computer Science - Artificial Intelligence</searchLink><br /><searchLink fieldCode="DE" term="%22Computer+Science+-+Machine+Learning%22">Computer Science - Machine Learning</searchLink> – Name: Abstract Label: Description Group: Ab Data: Large Language Models (LLMs) perform impressively well in various applications. However, the potential for misuse of these models in activities such as plagiarism, generating fake news, and spamming has raised concern about their responsible use. Consequently, the reliable detection of AI-generated text has become a critical area of research. AI text detectors have shown to be effective under their specific settings. In this paper, we stress-test the robustness of these AI text detectors in the presence of an attacker. We introduce recursive paraphrasing attack to stress test a wide range of detection schemes, including the ones using the watermarking as well as neural network-based detectors, zero shot classifiers, and retrieval-based detectors. Our experiments conducted on passages, each approximately 300 tokens long, reveal the varying sensitivities of these detectors to our attacks. Our findings indicate that while our recursive paraphrasing method can significantly reduce detection rates, it only slightly degrades text quality in many cases, highlighting potential vulnerabilities in current detection systems in the presence of an attacker. Additionally, we investigate the susceptibility of watermarked LLMs to spoofing attacks aimed at misclassifying human-written text as AI-generated. We demonstrate that an attacker can infer hidden AI text signatures without white-box access to the detection method, potentially leading to reputational risks for LLM developers. Finally, we provide a theoretical framework connecting the AUROC of the best possible detector to the Total Variation distance between human and AI text distributions. This analysis offers insights into the fundamental challenges of reliable detection as language models continue to advance. Our code is publicly available at https://github.com/vinusankars/Reliability-of-AI-text-detectors.<br />Comment: Published in Transactions on Machine Learning Research (TMLR) – Name: TypeDocument Label: Document Type Group: TypDoc Data: Working Paper – Name: URL Label: Access URL Group: URL Data: <link linkTarget="URL" linkTerm="http://arxiv.org/abs/2303.11156" linkWindow="_blank">http://arxiv.org/abs/2303.11156</link> – Name: AN Label: Accession Number Group: ID Data: edsarx.2303.11156
PLink	https://login.libproxy.scu.edu/login?url=https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&scope=site&db=edsarx&AN=edsarx.2303.11156
RecordInfo	BibRecord: BibEntity: Subjects: – SubjectFull: Computer Science - Computation and Language Type: general – SubjectFull: Computer Science - Artificial Intelligence Type: general – SubjectFull: Computer Science - Machine Learning Type: general Titles: – TitleFull: Can AI-Generated Text be Reliably Detected? Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Sadasivan, Vinu Sankar – PersonEntity: Name: NameFull: Kumar, Aounon – PersonEntity: Name: NameFull: Balasubramanian, Sriram – PersonEntity: Name: NameFull: Wang, Wenxiao – PersonEntity: Name: NameFull: Feizi, Soheil IsPartOfRelationships: – BibEntity: Dates: – D: 17 M: 03 Type: published Y: 2023
ResultId	1