The remainder of the thesis is organized as follow,chapter 2 explain the proposed approach, chapter 3 shows the related work, chapter 4 containsthe applied experiments and the related datasets.
And chapter 5 is the conclusion.1.4 Thesis OutlineLiveness Detection isan important research field in a biometric system. It’s become very active iniris and fingerprint recognition systems, and it still limited in facebiometric systems. Liveness detection detects whether the person in front ofthe camera is live (the real one ) or not live(spoofed one).LivenessDetection use varies approaches according to spoofing method as spoofing can bedone using face image, face image with the abilityto show eye blinking and mouth movement, 3D mask,video recorded of user’s face. 1.3 LivenessDetection Identificationmode: used in negative recognition which preventsthe user to use many identities, as in this systems the identification is doneusing one-to-many comparison to determine whether the user is true or it’sattacked Verification mode: used in positive recognition which aims to preventmany people from using the same identity, as in this systems the verificationis done using one-to-one comparison to determine whether the user is true orit’s attacked Biometrics systems use uniquefeatures of the user these features areanatomy and physiology ones as user face,hand geometry, user voice, iris or fingerprint.
These systems are commerciallyavailable today and are already in use. The biometrics systems used inidentification modes, as there are 2modes which are verification and identification modes.1.2 BiometricsThis research uses non-intrusive approach, this research tries to enhance OULU approach thatpresented in IJCB201 2 and it’s considered as one of the bestapproaches in this competitions as it achieveshigh performance in print attack dataset 3.OULU extract texture features and local features using LBP4, HOG5 and Gabor Wavelet6,But afterapplying OULU approach in other databases the performance decreases, so thisresearch reaches to generic parametersusing SIFT7 instead of HOG5 on Replay Attack8, CASIA9 and NUAA10Datasets and achieve better performance with the same SIFT parameters.
The intrusive approachdepends on user co-operation to detectthe life sign but In non-intrusive approach,extracting the life sign without any user co-operation by detecting the eyeblinking of the user, extracting sometexture features of the face which differs from the real face and spoofed one.Liveness detectiondetermines if a live person is present atthe time of capture or spoofed one.Liveness Detection can be done usinghardware devices that have the ability to look beyond the surface of the skinand can discriminate between the features of live skin and copies of thosefeatures in a fraction of a second. But this hardware solution is very expensiveand the improvement in such solution is very hard. The better solution achievedby using software applications, which can be Intrusive or non-intrusive.1.1 State Of The Art But the most commonused face recognition systems are easy to be attacked as listed in the NationalVulnerability Database of the National Institute of Standards and Technology(NIST) in the US as mentioned in 1, one can spoof a face recognition systemby presenting a photograph, a video, a mask or a 3D model of a targeted personin front of the camera.
Also, there are a lot of other methods for an attack like using plastic surgery or make-up or using the photographs which considered as the most common sources of spoofing attacks because one caneasily download and capture facial images. Liveness detection research has avital role in anti-spoofing of face recognition systems and biometrics systemsat all. As it detects the live person in front of the camera or it’s spoofedone. In an increasinglydigital world, protecting confidential information is becoming more difficult.
Traditional passwords and keys no longer provide enough security to ensure thatdata is kept out of the hands of hackers and unauthorized individuals. This iswhere biometric security can transform the technology sector. Biometricauthentication devices use unique traits or behavioralcharacteristics, such as fingerprint, hand, face, eye (iris or retina), and voice recognition, to authenticate access to electronic assets.Because biometric information is unique to each person as various biometricssystems have been developed around unique characteristics of individuals. Theprobability of 2 people sharing the same biometric data is virtually nil,cannot be shared because a biometric property is an intrinsic property of anindividual, also it can’t be lost only in case of serious accident.