NDSS 2025 – EMIRIS: Eavesdropping On Iris Information Via Electromagnetic Side Channel

Why It Matters

EMIRIS shows that biometric systems thought secure can be compromised via passive electromagnetic side‑channels, forcing a rethink of sensor design and deployment safeguards.

Key Takeaways

• EM emissions leak iris data from NIR sensors
• EMIRIS reconstructs iris images with SSIM 0.511
• Spoof success rate exceeds 53% on 3k samples
• Diffusion model solves linear inverse problem for denoising
• Highlights privacy risk for biometric authentication systems

Pulse Analysis

Iris recognition remains a cornerstone of high‑assurance authentication, prized for its uniqueness and resistance to forgery. However, the reliance on near‑infrared (NIR) illumination introduces an unexpected emission profile: the sensor’s digital signal transmission radiates electromagnetic (EM) waves that correlate with the captured iris texture. This side‑channel, long overlooked in biometric threat models, provides an attacker with a covert data source that can be harvested without physical contact, expanding the attack surface beyond traditional spoofing methods such as printed eyes or contact lenses.

The EMIRIS framework capitalizes on this leakage by first decoding the proprietary transmission protocol of commercial NIR sensors. Once the raw EM waveform is captured, the researchers model the reconstruction as a linear inverse problem, employing a tailored diffusion model to denoise and restore fine iris details. The approach yields reconstructed images with an average SSIM of 0.511 and a low FID of 7.25, metrics that indicate perceptible similarity to genuine irises. Crucially, when fed into a standard iris matcher, these synthetic irises achieve a 53.47% spoofing success rate across a diverse dataset of more than 3,000 samples, underscoring the practical viability of the attack.

The implications for enterprises and governments deploying iris scanners are profound. Existing security assessments must now account for EM side‑channel emissions, prompting hardware redesigns that incorporate shielding, randomized transmission timing, or encrypted data paths. Regulators may also tighten standards for biometric devices, demanding electromagnetic compliance testing alongside traditional performance metrics. As organizations increasingly adopt contactless authentication, the EMIRIS discovery serves as a timely reminder that even the most trusted biometric modalities can harbor hidden vulnerabilities, urging a proactive, defense‑in‑depth strategy.