Copertina

Trojans in Artificial Intelligence (TrojAI) Final Report

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Reese, Kristopher W., Kulp-McDowall, Taylor, Majurski, Michael, Blattner, Tim, Juba, Derek, Bajcsy, Peter, Cardone, Antonio, Dessauw, Philippe, Dima, Alden, Kearsley, Anthony J., Kleczynski, Melinda, Vasanth, Joel, Keyrouz, Walid, Ashcraft, Chace, Fendley, Neil, Staley, Ted, Stout, Trevor, Carney, Josh, Canal, Greg, Redman, Will, Schmidt, Aurora, Hickert, Cameron, Paul, William, Markowitz, Jared, Drenkow, Nathan, Shriver, David, Connor, Marissa, Grimes, Keltin, Christiani, Marco, Moore, Hayden, Widjaja, Jordan, Gabert, Kasimir, Balakrishnan, Uma, Gundimada, Satyanadh, Jacobellis, John, Lakkur, Sandya, Leung, Vitus, Roose, Jon, Battaglino, Casey, Koushanfar, Farinaz, Fields, Greg, Gu, Xihe, Jandali, Yaman, Zhang, Xinqiao, Javidi, Tara, Vartak, Akash, Oates, Tim, Erichson, Ben, Mahoney, Michael, Izmailov, Rauf, Zhang, Xiangyu, Shen, Guangyu, Cheng, Siyuan, Ma, Shiqing, Wang, XiaoFeng, Tang, Haixu, Tang, Di, Chen, Xiaoyi, Wang, Zihao, Zhu, Rui, Jha, Susmit, Lin, Xiao, Acharya, Manoj, Zhou, Weichao, Fu, Feisi, Kiourti, Panagiota, Wang, Chenyu, Guo, Zijian, Ahmad, H M Sabbir, Li, Wenchao, Chen, Chao
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Cryptography and Security Artificial Intelligence Machine Learning
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

_version_	1866914358260400128
author	Reese, Kristopher W. Kulp-McDowall, Taylor Majurski, Michael Blattner, Tim Juba, Derek Bajcsy, Peter Cardone, Antonio Dessauw, Philippe Dima, Alden Kearsley, Anthony J. Kleczynski, Melinda Vasanth, Joel Keyrouz, Walid Ashcraft, Chace Fendley, Neil Staley, Ted Stout, Trevor Carney, Josh Canal, Greg Redman, Will Schmidt, Aurora Hickert, Cameron Paul, William Markowitz, Jared Drenkow, Nathan Shriver, David Connor, Marissa Grimes, Keltin Christiani, Marco Moore, Hayden Widjaja, Jordan Gabert, Kasimir Balakrishnan, Uma Gundimada, Satyanadh Jacobellis, John Lakkur, Sandya Leung, Vitus Roose, Jon Battaglino, Casey Koushanfar, Farinaz Fields, Greg Gu, Xihe Jandali, Yaman Zhang, Xinqiao Javidi, Tara Vartak, Akash Oates, Tim Erichson, Ben Mahoney, Michael Izmailov, Rauf Zhang, Xiangyu Shen, Guangyu Cheng, Siyuan Ma, Shiqing Wang, XiaoFeng Tang, Haixu Tang, Di Chen, Xiaoyi Wang, Zihao Zhu, Rui Jha, Susmit Lin, Xiao Acharya, Manoj Zhou, Weichao Fu, Feisi Kiourti, Panagiota Wang, Chenyu Guo, Zijian Ahmad, H M Sabbir Li, Wenchao Chen, Chao
author_facet	Reese, Kristopher W. Kulp-McDowall, Taylor Majurski, Michael Blattner, Tim Juba, Derek Bajcsy, Peter Cardone, Antonio Dessauw, Philippe Dima, Alden Kearsley, Anthony J. Kleczynski, Melinda Vasanth, Joel Keyrouz, Walid Ashcraft, Chace Fendley, Neil Staley, Ted Stout, Trevor Carney, Josh Canal, Greg Redman, Will Schmidt, Aurora Hickert, Cameron Paul, William Markowitz, Jared Drenkow, Nathan Shriver, David Connor, Marissa Grimes, Keltin Christiani, Marco Moore, Hayden Widjaja, Jordan Gabert, Kasimir Balakrishnan, Uma Gundimada, Satyanadh Jacobellis, John Lakkur, Sandya Leung, Vitus Roose, Jon Battaglino, Casey Koushanfar, Farinaz Fields, Greg Gu, Xihe Jandali, Yaman Zhang, Xinqiao Javidi, Tara Vartak, Akash Oates, Tim Erichson, Ben Mahoney, Michael Izmailov, Rauf Zhang, Xiangyu Shen, Guangyu Cheng, Siyuan Ma, Shiqing Wang, XiaoFeng Tang, Haixu Tang, Di Chen, Xiaoyi Wang, Zihao Zhu, Rui Jha, Susmit Lin, Xiao Acharya, Manoj Zhou, Weichao Fu, Feisi Kiourti, Panagiota Wang, Chenyu Guo, Zijian Ahmad, H M Sabbir Li, Wenchao Chen, Chao
contents	The Intelligence Advanced Research Projects Activity (IARPA) launched the TrojAI program to confront an emerging vulnerability in modern artificial intelligence: the threat of AI Trojans. These AI trojans are malicious, hidden backdoors intentionally embedded within an AI model that can cause a system to fail in unexpected ways, or allow a malicious actor to hijack the AI model at will. This multi-year initiative helped to map out the complex nature of the threat, pioneered foundational detection methods, and identified unsolved challenges that require ongoing attention by the burgeoning AI security field. This report synthesizes the program's key findings, including methodologies for detection through weight analysis and trigger inversion, as well as approaches for mitigating Trojan risks in deployed models. Comprehensive test and evaluation results highlight detector performance, sensitivity, and the prevalence of "natural" Trojans. The report concludes with lessons learned and recommendations for advancing AI security research.
format	Preprint
id	arxiv_https___arxiv_org_abs_2602_07152
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	Trojans in Artificial Intelligence (TrojAI) Final Report Reese, Kristopher W. Kulp-McDowall, Taylor Majurski, Michael Blattner, Tim Juba, Derek Bajcsy, Peter Cardone, Antonio Dessauw, Philippe Dima, Alden Kearsley, Anthony J. Kleczynski, Melinda Vasanth, Joel Keyrouz, Walid Ashcraft, Chace Fendley, Neil Staley, Ted Stout, Trevor Carney, Josh Canal, Greg Redman, Will Schmidt, Aurora Hickert, Cameron Paul, William Markowitz, Jared Drenkow, Nathan Shriver, David Connor, Marissa Grimes, Keltin Christiani, Marco Moore, Hayden Widjaja, Jordan Gabert, Kasimir Balakrishnan, Uma Gundimada, Satyanadh Jacobellis, John Lakkur, Sandya Leung, Vitus Roose, Jon Battaglino, Casey Koushanfar, Farinaz Fields, Greg Gu, Xihe Jandali, Yaman Zhang, Xinqiao Javidi, Tara Vartak, Akash Oates, Tim Erichson, Ben Mahoney, Michael Izmailov, Rauf Zhang, Xiangyu Shen, Guangyu Cheng, Siyuan Ma, Shiqing Wang, XiaoFeng Tang, Haixu Tang, Di Chen, Xiaoyi Wang, Zihao Zhu, Rui Jha, Susmit Lin, Xiao Acharya, Manoj Zhou, Weichao Fu, Feisi Kiourti, Panagiota Wang, Chenyu Guo, Zijian Ahmad, H M Sabbir Li, Wenchao Chen, Chao Cryptography and Security Artificial Intelligence Machine Learning The Intelligence Advanced Research Projects Activity (IARPA) launched the TrojAI program to confront an emerging vulnerability in modern artificial intelligence: the threat of AI Trojans. These AI trojans are malicious, hidden backdoors intentionally embedded within an AI model that can cause a system to fail in unexpected ways, or allow a malicious actor to hijack the AI model at will. This multi-year initiative helped to map out the complex nature of the threat, pioneered foundational detection methods, and identified unsolved challenges that require ongoing attention by the burgeoning AI security field. This report synthesizes the program's key findings, including methodologies for detection through weight analysis and trigger inversion, as well as approaches for mitigating Trojan risks in deployed models. Comprehensive test and evaluation results highlight detector performance, sensitivity, and the prevalence of "natural" Trojans. The report concludes with lessons learned and recommendations for advancing AI security research.
title	Trojans in Artificial Intelligence (TrojAI) Final Report
topic	Cryptography and Security Artificial Intelligence Machine Learning
url	https://arxiv.org/abs/2602.07152