Abstract
While various methods have been proposed to explain AI models, the trustworthiness of the generated explanation received little examination. This paper reveals that such explanations could be vulnerable to subtle perturbations on the input and generate misleading results. On the public CheXpert dataset, we demonstrate that specially designed adversarial perturbations can easily tamper saliency maps towards the desired explanations while preserving the original model predictions. AI researchers, practitioners, and authoritative agencies in the medical domain should use caution when explaining AI models because such an explanation could be irrelevant, misleading, and even adversarially manipulated without changing the model output. AI researchers, practitioners, and authoritative agencies in the medical domain should use caution when explaining AI models because such an explanation could be irrelevant, misleading, and even adversarially manipulated without changing the model output.
Competing Interest Statement
The authors have declared no competing interest.
Funding Statement
NSF CAREER 2046708
Author Declarations
I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained.
Yes
I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals.
Yes
I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance).
Yes
I have followed all appropriate research reporting guidelines and uploaded the relevant EQUATOR Network research reporting checklist(s) and other pertinent material as supplementary files, if applicable.
Yes
Data Availability
All data produced are available online at