This article presents a novel threat model in which attackers use webcams to capture screen content reflected in a victim's glasses. The attack relies on three fundamental technical enablers: reflection pixel size, viewing angle, and optical signal-to-noise ratio (SNR). Through empirical testing and mathematical modeling, it demonstrates how even low-resolution webcams can extract readable text by applying advanced image enhancement techniques such as Multi Frame Super Resolution (MFSR).
Technical description: the reflection pixel size determines how much screen information is sampled by each camera pixel. The viewing angle conditions the geometry of the reflection and the apparent scale of the characters. The optical signal-to-noise ratio (SNR) controls the visibility of the content against optical and electronic noise. Combining these factors makes it possible to estimate minimum resolution and SNR thresholds for text to be reconstructible using super-resolution and image restoration techniques.
Methodology and findings: using experiments with different webcams and lighting conditions, video sequences of the reflection on lenses were recorded. The processing applied multi-frame registration, deconvolution filters, and deep learning models for MFSR. The results show that when the reflection covers at least a critical number of pixels per character and the SNR exceeds a threshold, partial or complete text reading is feasible even with commercial low-resolution sensors.
Security implications: the continuous improvement of webcam hardware and the proliferation of lenses with reflective surfaces increase the risk of information leakage through reflection. Small, seemingly innocuous optical details, such as lens curvature or the presence of glints, can facilitate privacy attacks that exfiltrate passwords, document fragments, or sensitive data displayed on screen.
Practical countermeasures: reduce physical risks by using anti-reflective filters and lenses with matte treatment, adjust the position of the screen and the user to minimize angles that generate reflections, control ambient lighting to lower the reflection's SNR, and use privacy screen protectors. On the digital front, it is advisable to apply visual obfuscation of sensitive data, temporary masks on critical interfaces, and security policies that limit the access of capture devices to sensitive environments.
Detection and resilience: artificial intelligence-based solutions can monitor unusual camera signals and detect attempts to capture screens via reflection. Detection models can combine device metadata, motion pattern analysis, and optical heuristics to block or alert on potential exfiltration. Additionally, font and spacing design techniques can reduce readable reconstruction by MFSR, mitigating the risk without affecting usability.
How Q2BSTUDIO can help: at Q2BSTUDIO, we are a custom software and application development company specialized in artificial intelligence and cybersecurity. We offer comprehensive services including security audits, design of AI-based detection solutions for businesses, custom software development, and custom applications with a privacy focus. We implement secure cloud architectures with AWS and Azure cloud services, develop AI agents, and business intelligence solutions to visualize and protect sensitive assets with tools such as Power BI.
Our services: implementation of custom MFSR models and detection algorithms, development of security policies and training, deployment in cloud environments with AWS and Azure cloud services, integration of artificial intelligence solutions to automate incident response, and development of dashboards with Power BI to monitor risk indicators. All backed by experience in custom software and cybersecurity to minimize exposure to reflection-based attacks.
Final recommendations: assess the optical exposure surface in corporate environments, apply physical and logical measures to reduce the reflection's SNR, incorporate advanced detection through artificial intelligence, and rely on technology partners that offer custom software and implementation on AWS and Azure cloud services. Q2BSTUDIO can assist in designing mitigation strategies, developing business intelligence solutions, and implementing AI agents tailored to each company's needs.
Conclusion: the threat model of reflection attacks on glasses reveals that physical and optical vulnerabilities combined with advanced image processing techniques can enable significant information leaks. Given the evolution of hardware and AI capabilities, it is essential to adopt a holistic cybersecurity strategy that includes physical controls, custom software, artificial intelligence for detection and response, and secure deployment on AWS and Azure cloud services. Q2BSTUDIO is prepared to design and implement custom solutions that protect your organization's critical information.
