MIDV-DynAttack

MIDV‑DynAttack is the dataset for the paper "Verification of Dynamic Holographic Behavior in Identity Documents" accepted at the 19th International Conference on Document Analysis and Recognition (ICDAR 2025). This dataset addresses the remote verification of the authenticity of Optically Variable Devices (commonly known as holograms) on identity documents. Typically placed over the cardholder’s photo, these devices provide strong and easily verifiable security for human inspection but pose challenges for automated verification. Existing approaches easily cover static frauds (e.g., paper photocopy) and can be evaluated for such, but their capacity to detect real, dynamic fraud cases (e.g., handcrafted hologram) has not been evaluated to date because of the lack of public datasets. Furthermore, many systems are trained to detect known attack types, and few can generalize to new, unseen attacks. MIDV‑DynAttack is a new public dataset that extends the existing MIDV‑Holo dataset with realistic static and dynamic attacks against identity document specimens, tripling the number of attack samples compared to the original dataset. It is intended as a test‑only set of previously unseen attacks for evaluating the generalization and robustness of hologram verification methods. This record provides: video.tar.gz: 1200 newly captured videos (raw sequences for MIDV‑DynAttack). tracking.tar.gz: document localization coordinates for each frame sampled at 15 fps. cropped.tar.gz: rectified Regions of Interest (ROIs) containing part of the face and the main hologram, computed using the document localization. Directory layout MIDV‑DynAttack Static-template attacks fraud/natural_light_paper/ Template without lamination (Natural Light - Paper) fraud/natural_light_plastic_low/ Template with lamination, minimal reflections (Natural Light - Plastic, minimal reflections) fraud/natural_light_plastic_hi/ Template with lamination, deliberate reflections (Natural Light - Plastic, multiple light sources) fraud/led_light_plastic/ Laminated, illuminated by multi-color LED ring (LED-light - Plastic) fraud/laser_light_plastic/ Laminated, red laser moving over the document (Laser light - Plastic) fraud/swap_two/ Templates without lamination, swapping between two printed documents fraud/swap_three/ Templates without lamination, swapping between three printed documents Dynamic-template attacks fraud/plain_holo/ Homogeneous holographic overlay (continuous diffraction) fraud/leaf_holo/ Overlay with polygonal “leaf” patterns fraud/double_sticker/ Flower-shaped laser-film element + a “genuine” sticker (non-overlapping ROI) fraud/complete_mask/ Rough mask of the hologram applied over the entire printed document fraud/star_world/ Combination of a main flower shape and a world-like hologram pattern MIDV‑Holo (reference dataset) origins/ fraud/photo_holo_copy/ fraud/pseudo_holo_copy/ fraud/copy_without_holo/ fraud/photo_replacement/ Note: cropped.tar.gz and tracking.tar.gz also contains the MIDV-Holo samples.MIDV‑DynAttack是为收录于第19届国际文档分析与识别会议（ICDAR 2025）的论文《验证身份证件的动态全息行为》所构建的数据集。 本数据集聚焦身份证件上光学可变装置（Optically Variable Devices，俗称全息图）的远程真实性验证问题。这类装置通常设置在持卡人照片区域上方，可为人工核验提供可靠且易于验证的安全防护，但却给自动化核验带来了挑战。现有方法可轻松应对静态伪造场景（如纸质复印）并针对此类场景开展评估，但截至目前，由于缺乏公开数据集，相关方法检测真实动态伪造场景（如手工制作的全息图）的能力尚未得到验证。此外，多数系统仅针对已知攻击类型进行训练，极少能够泛化至未见过的新型攻击。 MIDV‑DynAttack是全新的公开数据集，它在现有MIDV‑Holo数据集的基础上，新增了针对身份证件样本的写实静态与动态攻击，攻击样本数量较原始数据集增加了两倍。该数据集仅作为测试集，包含未见过的攻击样本，用于评估全息图验证方法的泛化能力与鲁棒性。 本数据集包含以下内容： - video.tar.gz：1200段新采集的视频（MIDV‑DynAttack的原始序列数据） - tracking.tar.gz：以15帧/秒采样的每一帧对应的证件定位坐标 - cropped.tar.gz：经校正的感兴趣区域（Regions of Interest，ROIs），包含部分人脸与主全息图区域，通过证件定位信息计算得到 目录布局 MIDV‑DynAttack ├─ 静态模板攻击 │ ├─ fraud/natural_light_paper/：无覆膜模板（自然光-纸质） │ ├─ fraud/natural_light_plastic_low/：带覆膜模板，反射极弱（自然光-塑料，低反射） │ ├─ fraud/natural_light_plastic_hi/：带覆膜模板，刻意制造反射（自然光-塑料，多光源） │ ├─ fraud/led_light_plastic/：带覆膜模板，由多色LED环照明（LED光源-塑料） │ ├─ fraud/laser_light_plastic/：带覆膜模板，红色激光在证件表面移动（激光光源-塑料） │ ├─ fraud/swap_two/：无覆膜模板，在两份打印文档间切换 │ └─ fraud/swap_three/：无覆膜模板，在三份打印文档间切换 ├─ 动态模板攻击 │ ├─ fraud/plain_holo/：均匀全息覆盖层（连续衍射） │ ├─ fraud/leaf_holo/：带有多边形“叶片”图案的覆盖层 │ ├─ fraud/double_sticker/：花形激光薄膜元件+“正版”贴纸（非重叠ROI） │ ├─ fraud/complete_mask/：在整个打印文档表面覆盖全息图的粗糙蒙版 │ └─ fraud/star_world/：主花形图案与类世界全息图案的组合 └─ MIDV‑Holo（参考数据集） ├─ origins/ ├─ fraud/photo_holo_copy/ ├─ fraud/pseudo_holo_copy/ ├─ fraud/copy_without_holo/ └─ fraud/photo_replacement/ 注意：cropped.tar.gz与tracking.tar.gz同样包含MIDV-Holo的样本。