Arslan Ali, Andrea Migliorati, Tiziano Bianchi, Enrico Magli
{"title":"基于深度学习和随机投影的安全人脸验证可取消模板","authors":"Arslan Ali, Andrea Migliorati, Tiziano Bianchi, Enrico Magli","doi":"10.1186/s13635-023-00147-y","DOIUrl":null,"url":null,"abstract":"Recently, biometric recognition has become a significant field of research. The concept of cancelable biometrics (CB) has been introduced to address security concerns related to the handling of sensitive data. In this paper, we address unconstrained face verification by proposing a deep cancelable framework called BiometricNet+ that employs random projections (RP) to conceal face images and compressive sensing (CS) to reconstruct measurements in the original domain. Our lightweight design enforces the properties of unlinkability, revocability, and non-invertibility of the templates while preserving face recognition accuracy. We compare facial features by learning a regularized metric: at training time, we jointly learn facial features and the metric such that matching and non-matching pairs are mapped onto latent target distributions; then, for biometric verification, features are randomly projected via random matrices changed at every enrollment and query and reconstructed before the latent space mapping is computed. We assess the face recognition accuracy of our framework on challenging datasets such as LFW, CALFW, CPLFW, AgeDB, YTF, CFP, and RFW, showing notable improvements over state-of-the-art techniques while meeting the criteria for secure cancelable template design. Since our method requires no fine-tuning of the learned features, it can be applied to pre-trained networks to increase sensitive data protection.","PeriodicalId":46070,"journal":{"name":"EURASIP Journal on Information Security","volume":null,"pages":null},"PeriodicalIF":2.5000,"publicationDate":"2024-03-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Cancelable templates for secure face verification based on deep learning and random projections\",\"authors\":\"Arslan Ali, Andrea Migliorati, Tiziano Bianchi, Enrico Magli\",\"doi\":\"10.1186/s13635-023-00147-y\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Recently, biometric recognition has become a significant field of research. The concept of cancelable biometrics (CB) has been introduced to address security concerns related to the handling of sensitive data. In this paper, we address unconstrained face verification by proposing a deep cancelable framework called BiometricNet+ that employs random projections (RP) to conceal face images and compressive sensing (CS) to reconstruct measurements in the original domain. Our lightweight design enforces the properties of unlinkability, revocability, and non-invertibility of the templates while preserving face recognition accuracy. We compare facial features by learning a regularized metric: at training time, we jointly learn facial features and the metric such that matching and non-matching pairs are mapped onto latent target distributions; then, for biometric verification, features are randomly projected via random matrices changed at every enrollment and query and reconstructed before the latent space mapping is computed. We assess the face recognition accuracy of our framework on challenging datasets such as LFW, CALFW, CPLFW, AgeDB, YTF, CFP, and RFW, showing notable improvements over state-of-the-art techniques while meeting the criteria for secure cancelable template design. Since our method requires no fine-tuning of the learned features, it can be applied to pre-trained networks to increase sensitive data protection.\",\"PeriodicalId\":46070,\"journal\":{\"name\":\"EURASIP Journal on Information Security\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":2.5000,\"publicationDate\":\"2024-03-08\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"EURASIP Journal on Information Security\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1186/s13635-023-00147-y\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"COMPUTER SCIENCE, INFORMATION SYSTEMS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"EURASIP Journal on Information Security","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1186/s13635-023-00147-y","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"COMPUTER SCIENCE, INFORMATION SYSTEMS","Score":null,"Total":0}
Cancelable templates for secure face verification based on deep learning and random projections
Recently, biometric recognition has become a significant field of research. The concept of cancelable biometrics (CB) has been introduced to address security concerns related to the handling of sensitive data. In this paper, we address unconstrained face verification by proposing a deep cancelable framework called BiometricNet+ that employs random projections (RP) to conceal face images and compressive sensing (CS) to reconstruct measurements in the original domain. Our lightweight design enforces the properties of unlinkability, revocability, and non-invertibility of the templates while preserving face recognition accuracy. We compare facial features by learning a regularized metric: at training time, we jointly learn facial features and the metric such that matching and non-matching pairs are mapped onto latent target distributions; then, for biometric verification, features are randomly projected via random matrices changed at every enrollment and query and reconstructed before the latent space mapping is computed. We assess the face recognition accuracy of our framework on challenging datasets such as LFW, CALFW, CPLFW, AgeDB, YTF, CFP, and RFW, showing notable improvements over state-of-the-art techniques while meeting the criteria for secure cancelable template design. Since our method requires no fine-tuning of the learned features, it can be applied to pre-trained networks to increase sensitive data protection.
期刊介绍:
The overall goal of the EURASIP Journal on Information Security, sponsored by the European Association for Signal Processing (EURASIP), is to bring together researchers and practitioners dealing with the general field of information security, with a particular emphasis on the use of signal processing tools in adversarial environments. As such, it addresses all works whereby security is achieved through a combination of techniques from cryptography, computer security, machine learning and multimedia signal processing. Application domains lie, for example, in secure storage, retrieval and tracking of multimedia data, secure outsourcing of computations, forgery detection of multimedia data, or secure use of biometrics. The journal also welcomes survey papers that give the reader a gentle introduction to one of the topics covered as well as papers that report large-scale experimental evaluations of existing techniques. Pure cryptographic papers are outside the scope of the journal. Topics relevant to the journal include, but are not limited to: • Multimedia security primitives (such digital watermarking, perceptual hashing, multimedia authentictaion) • Steganography and Steganalysis • Fingerprinting and traitor tracing • Joint signal processing and encryption, signal processing in the encrypted domain, applied cryptography • Biometrics (fusion, multimodal biometrics, protocols, security issues) • Digital forensics • Multimedia signal processing approaches tailored towards adversarial environments • Machine learning in adversarial environments • Digital Rights Management • Network security (such as physical layer security, intrusion detection) • Hardware security, Physical Unclonable Functions • Privacy-Enhancing Technologies for multimedia data • Private data analysis, security in outsourced computations, cloud privacy