face-recognition-transfer-attacks

A public repository of transfer-based adversarial attack implementations adapted for CNN-based face recognition / face verification systems.

This repository focuses on attack code, usage instructions, paper references, and contributor attribution. It does not aim to be a public benchmark leaderboard. The implementations here were adapted in a shared face-verification pipeline and are intended to help researchers and students quickly reuse, inspect, and extend transfer attacks in the face-recognition setting.

Scope

• CNN-oriented face-recognition attack adaptations
• transfer-based adversarial attacks for embedding-based face verification
• student-contributed implementations consolidated into one reusable codebase
• per-attack documentation with original paper references and contributor credit

Important note

Some attacks in this repository were originally proposed for generic image classification or Vision Transformer settings. The code included here reflects CNN-side face-verification adaptations, not necessarily exact official reproductions of the original paper implementations.

To avoid ambiguity, this repository distinguishes between:

• original paper authors, who proposed the method in the literature, and
• repository implementation contributors, who adapted or integrated a version of the attack for this CNN face-verification codebase.

Repository layout

• core/transfer_attack_core.py: shared attack implementations and registry
• experiments/generate_adversarial_examples.py: generic batch generation script
• attacks/: per-attack notes, references, contributor information, and student slide PDFs
• docs/usage.md: how to run the code
• CONTRIBUTORS.md: attack-wise contributor list

Attack catalog

Attack	Original paper / source	Original authors	Repository implementation / adaptation
PGD	Towards Deep Learning Models Resistant to Adversarial Attacks (2018)	Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, Adrian Vladu	Shared repository implementation integrated in the CNN face-verification pipeline by Vishaka from Delhi University
MI_FGSM	Boosting Adversarial Attacks with Momentum (CVPR 2018)	Yinpeng Dong, Fangzhou Liao, Tianyu Pang, Hang Su, Jun Zhu, Xiaolin Hu, Jianguo Li	Shared repository implementation integrated in the CNN face-verification pipeline by Vishaka from Delhi University
TI_FGSM	Evading Defenses to Transferable Adversarial Examples by Translation-Invariant Attacks (CVPR 2019)	Yinpeng Dong, Tianyu Pang, Hang Su, Jun Zhu	Shared repository implementation integrated in the CNN face-verification pipeline by Ananya Jain from Jaypee Institute of Information Technology, Noida
SI_NI_FGSM	Nesterov Accelerated Gradient and Scale Invariance for Adversarial Attacks (2020)	Jiadong Lin, Chuanbai Xiao, Lei Yang, Xiaoyuan Mao, Tian Xia, Tong He, Wei Yang	Shared repository implementation integrated in the CNN face-verification pipeline by Vishaka from Delhi University
MI_ADMIX_DI_TI	Admix (ICCV 2021) with DI/TI-style transfer components	Xiaosen Wang, Jiefeng Chen, Si'wei Lyu, Yisen Wang	Shared repository implementation integrated in the CNN face-verification pipeline by Ananya Jain from Jaypee Institute of Information Technology, Noida
MIG	Transferable Adversarial Attack for Both Vision Transformers and Convolutional Networks via Momentum Integrated Gradients (ICCV 2023)	Wenshuo Ma, Yidong Li, Xiaofeng Jia, Wei Xu	CNN face-verification adaptation contributed by Lakshita Sharma from Bhagwan Parshuram Institute of Technology
ADAMSI_FGM	On the Convergence of an Adaptive Momentum Method for Adversarial Attacks (AAAI 2024)	Yuxiao Long, Xuanyu Yi, Jinglun Xu, Yuan Wu, Xiaosen Wang	CNN face-verification adaptation contributed by Bhumika Singh from SRM Institute of Science and Technology
BPA_CNN	Rethinking the Backward Propagation for Adversarial Transferability (NeurIPS 2023)	Xiaosen Wang, Kangheng Tong, Kun He	CNN-side face-verification adaptation contributed by Om Singh Rawat from IIT Delhi
PGN	Boosting Adversarial Transferability by Achieving Flat Local Maxima (NeurIPS 2023)	Yao Ge, Yao Xiao, Weijie Huang, Xiaosen Wang, Yong Liu, Bo Han, Jiang Bian, Chunhua Shen	CNN face-verification adaptation contributed by Chidroopa Kanaparthy from Maharaja Agrasen Institute of Technology (MAIT)
BSR	Boosting Adversarial Transferability by Block Shuffle and Rotation (CVPR 2024)	Kunyu Wang, Xuanran He, Wenxuan Wang, Xiaosen Wang	Face-verification adaptation contributed by Chirag Sharma from IIIT Vadodara
DECOWA	Boosting Adversarial Transferability across Model Genus by Deformation-Constrained Warping (AAAI 2024)	Jiayi Lin, Chuanbai Xiao, Chao Ma, Jie Zhang, Qiong Cao, Xiaosen Wang	Face-verification adaptation contributed by Om Singh Rawat from IIT Delhi
SIA_MI_TI	Structure Invariant Transformation for better Adversarial Transferability (ICCV 2023)	Xiaosen Wang, Zeliang Zhang, Jianping Zhang	CNN face-verification adaptation contributed by Janhavi Kishor from SRM University
SIA	Structure Invariant Transformation for better Adversarial Transferability (ICCV 2023)	Xiaosen Wang, Zeliang Zhang, Jianping Zhang	CNN face-verification adaptation contributed by Aditi Raj from IGDTUW
OPS	Boosting Adversarial Transferability through Augmentation in Hypothesis Space (CVPR 2025)	Yu Guo, Weiquan Liu, Qingshan Xu, Shijun Zheng, Shujun Huang, Yu Zang, Siqi Shen, Chenglu Wen, Cheng Wang	Face-verification adaptation contributed by Kkartik Aggarwal from Delhi Technological University (DTU)
ATT_CNN	Boosting the Transferability of Adversarial Attack on Vision Transformer with Adaptive Token Tuning (NeurIPS 2024)	Di Ming, Peng Ren, Yunlong Wang, Xin Feng	CNN-side adaptation inspired by the original ATT method, contributed by Keshav Raj from IIIT Delhi
ATT_CNN_PATCH	Boosting the Transferability of Adversarial Attack on Vision Transformer with Adaptive Token Tuning (NeurIPS 2024)	Di Ming, Peng Ren, Yunlong Wang, Xin Feng	Separate CNN-side stochastic patch-masking adaptation inspired by ATT, contributed by Pratyush Kumar from KCC Institute of Technology and Management, A.K.T.U.
LI_BOOST_MI	Boosting the Local Invariance for Better Adversarial Transferability (arXiv 2025)	Bohan Liu, Xiaosen Wang	CNN face-verification adaptation contributed by Charushi from IGDTUW
GRA	Boosting Adversarial Transferability via Gradient Relevance Attack (ICCV 2023)	Yingwen Wu, Yinpeng Dong, Qin Wang, Jun Zhu, Xiaolin Hu	CNN face-verification adaptation contributed by Krish Bansal from Delhi Technological University (DTU)
IDAA	Student-submitted IDAA basis for CNN face verification	Not independently standardized in the submitted materials	CNN face-verification adaptation contributed by Arnav Asthana from IIIT Delhi
DPA_HMA	Improving the Transferability of Adversarial Attacks on Face Recognition with Diverse Parameters Augmentation (CVPR 2025)	Fengfan Zhou, Bangjie Yin, Hefei Ling, Qianyu Zhou, Wenxuan Wang	Face-recognition adaptation contributed by Kushal Khemka from Delhi Technological University (DTU)
DYNAMIC_MORPH	Student-submitted morph-style semantic face-region mixing basis for CNN face verification	Not independently standardized in the submitted materials	CNN face-verification adaptation contributed by Puneet Kumar from Indian Institute of Information Technology, Senapati, Manipur

Quick start

Install dependencies:

pip install -r requirements.txt

Generate adversarial images from a CSV of source-target pairs:

python experiments/generate_adversarial_examples.py \
  --input-csv /path/to/input_pairs.csv \
  --dataset-root /path/to/dataset_extractedfaces \
  --output-root /path/to/adv_outputs \
  --attacker-model Facenet512 \
  --attacks BSR,DPA_HMA,LI_BOOST_MI

For input CSV format and field descriptions, see docs/usage.md.

Acknowledgment

This repository consolidates student implementations created in a guided research setting. Each attack folder includes the contributor name, college affiliation, original paper reference, and an implementation note describing the adaptation.