DeepAware: Using Experiential Deepfake Simulations to Enhance Cybersecurity Awareness in Older Adults

Paper Title

DeepAware: Using Experiential Deepfake Simulations to Enhance Cybersecurity Awareness in Older Adults

Publication Info

• Topic area: Cybersecurity education for older adults using personalized deepfake simulations.
• Keywords: Deepfake scams, cybersecurity education, older adults, self-referential simulation, experiential learning, threat perception, coping strategies, digital literacy, AI-generated media, Protection Motivation Theory.

Background and Problem

• Problem / challenge: Older adults are particularly vulnerable to deepfake scams due to limited awareness, cognitive challenges, and lack of practical training. Existing educational approaches are abstract and fail to make threats feel personally relevant, leaving older adults unprepared to respond effectively.
• Significance: Deepfake scams exploit emotional urgency and social trust, posing serious risks to individuals and society. Addressing this issue is critical to improving digital safety and empowering older adults to protect themselves.
• Motivation and related work: Prior research has focused on technical detection methods and general digital literacy campaigns, which often fail to translate knowledge into protective behavior. Experiential learning and self-referential personalization have shown promise in enhancing engagement and retention but have not been applied to deepfake education for older adults.

Solution

• Proposed approach: DeepAware, a web-based platform that uses personalized deepfake simulations to educate older adults about deepfake scams. The system embeds users’ own faces and voices into simulated scenarios to enhance threat perception and coping efficacy.
• Novelty:
  1. Embedding learners’ own identities into deepfake scenarios to make threats personally relevant.
  2. Designing scaffolded learning stages that progress from conceptual education to personalized simulations and coping strategies.
  3. Providing actionable, concrete strategies for responding to deepfake scams.
  4. Evaluating the approach with older adults to assess its impact on knowledge, threat perception, and coping confidence.
• Procedure and key techniques:
  • Users upload a photo and voice recording, which are used to generate personalized deepfake simulations.
  • Two modules (Identity Theft and Fake News) present scenarios where the user’s likeness is manipulated.
  • A five-stage learning flow includes concept introduction, real-world examples, self-relevant simulations, coping strategies, and review quizzes.
  • An animated self-resembling guide narrates the program, enhancing engagement and reducing anxiety.

Results

• Concrete findings:
  • Significant improvements in deepfake knowledge (+1.0 on a 5-point scale, p = .002), perceived vulnerability (+0.91, p = .002), perceived severity (+0.71, p = .010), self-efficacy (+0.85, p = .003), and response efficacy (+0.90, p = .001).
  • High satisfaction (Median = 5.0) and perceived usefulness (Median = 4.5), with moderate ratings for deepfake realism (Median = 3.0).
• Advantage over baselines: Personalized simulations made threats feel immediate and tangible, addressing the abstract nature of traditional approaches. Scaffolded learning supported both threat appraisal and coping efficacy.
• Experiments / evaluation:
  • Single-group pre–post study with 21 older adults (Mean age = 65.8) in South Korea.
  • Mixed-methods evaluation included surveys, think-aloud protocols, and interviews.
  • Measures assessed deepfake knowledge, perceived threat, coping efficacy, and learning experience.
• Limitations and future work:
  • Sample skewed toward digitally engaged older adults; findings may not generalize to less experienced users.
  • Single-group design limits causal inference; future randomized controlled trials are needed.
  • Short-term evaluation; long-term retention and behavioral impact remain unclear.
  • Limited scenario diversity; future work should include more scam types and improve simulation realism.

Summary

DeepAware is a personalized, simulation-based platform designed to educate older adults about deepfake scams by embedding their own faces and voices into realistic scenarios. The system significantly improved participants’ knowledge, threat perception, and coping efficacy, demonstrating the potential of self-referential experiential learning. While participants found the approach engaging and actionable, challenges such as varied baseline knowledge and simulation realism highlight the need for adaptive and scalable designs. This work underscores the promise of personalized cybersecurity education to empower older adults against emerging AI-driven threats.