Work

Combating Misinformation in the Age of Deepfakes

Year: 2025

Type: article

Abstract: The study discusses the defense in relation to the problem of deepfakes using a mixed methodology by implementing technical assessment, randomized controlled tests (RCT), and interviews with practitioners. The best out-of-distribution performance was the multimodal fusion and late ensembling (ensemble AUC 0.979 in-domain, 0.944 out-of-domain) and high-resistance to post-processing methods. Combining prebunking with AI-generated labels led to a 31.4% reduction in the intent to share synthetic content compared to the control group in the RCT (n ≈ 900) and an apparent reduction in the perceived accuracy of fakes. Provenance and watermarks checks were found to be very precise, covering 23% of items. The practitioner responses mentioned clear labels, confidence displays, and audit processes as being of significant value. Deep fake threats can be efficiently addressed by using a layered defense mechanism, which consists of calibrated multimodal detection, interoperable provenance, and timely user education.

Source: Global Political Review

DOI: 10.31703/gpr.2025(x-iii).12

Authors: Muntaha Sami, Zarwah Nabil, Maryam Hashmi

Institutions: Minhaj University Lahore, University of the Punjab, International Islamic University, Islamabad

Language: English

Topic: Misinformation and Its Impacts

Subfield: Sociology and Political Science

Field: Social Sciences

Domain: Social Sciences

SDG: Peace, justice, and strong institutions

Open Access status: hybrid

OpenAlex ID: W4415113143

FWCI:

Cites:

Cited by:

Related to: