Avinash Kumar Singh

Face recognition with liveness detection using eye and mouth movement

The recent literature on face recognition technology discusses the issue of face spoofing which can bypass the authentication system by placing a photo/video/mask of the enrolled person in front of the camera. This problem could be minimized by detecting the liveness of the person. Therefore, in this paper, we propose a robust liveness detection scheme based on challenge and response method. The liveness module is added as extra layer of security before the face recognition module. The liveness module utilizes face macro features, especially eye and mouth movements in order to generate random challenges and observing the users response on account of this. The reliability of liveness module is tested by placing different types of spoofing attacks with various means, like using photograph, videos, etc. In all, five types of attacks have been taken care of and prevented by our system. Experimental results show that system is able to detect the liveness when subjected to all these attacks except the eye & mouth imposter attack. This attack is able to bypass the liveness test but it creates massive changes in face structure. Therefore resultant unrecognized or misclassified by the face recognition module. An experimental test conducted on 65 persons on university of Essex face database confirms that removal of eye and nose components results 75% misclassification.

A network based vulnerability scanner for detecting SQLI attacks in web applications

Today is the world of information era, where information is available on just our single click. Web applications are playing a magnificent role in this, every organizations are mapping their business from a room to the world with the help of these Web Apps. Web applications generally consist of a three tier architecture where database is in the third pole, which is the most valuable assets in any organization, as the adaptation of web applications are increases day by day, various attacks are possible against this. SQL injection is an attack in which an attacker directly compromises the database, thats why this is a most threatening attack. Various Vulnerability scanners has been proposed to deal with this, but none of them are able to detect SQLI completely, the existing tools have the accuracy ratio very less as well as they produce a high rate of false positive, apart from that all these tools take much time to scan. So here we are presenting a network based vulnerability scanner approach which provides a better coverage and with no false positive within a short span of time.

Face recognition using facial symmetry

Face is the most frequently used biometric trait after fingerprint. Its applicability made it popular in different areas such as Human Robot Interaction (HRI), Security Authentication, and Surveillance to name a few. Face recognition concept is based on two major blocks, training and testing. Usually training is done offline while testing is performed in real time scenario. As the size of the database increases, the recognition rate (time taken by system to recognize) increases. The rate of recognition is directly proportional to the size of the database and the dimension of the images. Human faces have the vertical symmetry; hence we utilized this feature and proposed a half way face recognition approach. Experimental verification on both the full faces and the half faces shows that half faces are also sufficient for recognizing the person. For verifying the efficiency of the approach, we have applied PCA (Principle Component Analysis) on both, the full faces and half faces, and have found that in both the cases, accuracy is almost same. But the recognition rate of half faces is just the half of the full faces.

NAO humanoid robot

The strength of the NAO humanoid robot is discussed with several challenges in the arena of human portrait and sketch drawing. These challenges include extracting feature points from the input image, defining these points with respect to end effector, finding the inverse kinematics solution and designing a visual feedback system. This paper mainly addresses the fundamental issue of defining a relationship between the points of the image plane and NAO end effector position. This relationship enables NAO to perceive points of image plane with respect to its body coordinate system. Three different techniques based on the principles of fundamental matrix, pseudo inverse and Artificial Neural Network based regression analysis are applied to handle the calibration difficulties on NAO robot. A comprehensive study on sample points collected from NAO end effector position and corresponding image points has been made to understand the effectiveness of each technique. The degree of performance ability of each technique has been measured using time complexity, and mean square error metrics. The paper proposed 3 novel techniques to calibrate humanoid camera plane to its end-effector position. The solution is generalized consider the experimental setup is same.Extensively error analysis and time complexity analysis is performed to evaluate each proposed technique.Proposed techniques are computationally sound and require only few points (4 points) to calibrate the system. The error involved in calibration is also very less.

Audio steganography using LSB encoding technique with increased capacity and bit error rate optimization

Staganography is the art and science of secret hiding. The secret message or plain text may be hidden in one various ways. The methods of cryptography render the message unintelligible to the outsider by various transformations of the text whereas the methods of steganography conceal the existence of the message. To conceal a secret message we need a wrapper or container as a host file. Different wrappers or host files or cover medium are used to hide the secret message e.g. image, audio, video, text. The work in this paper aims at enhancing the provision of audio steganography by introducing one LSB (Least Significant Bit) coding technique. We design a high bit rate LSB audio watermarking method that reduces embedding distortion of the host audio with increased capacity of secret text. By using standard and proposed algorithm, watermark bits are embedded into higher LSB layer, resulting in increased robustness against noise addition, which is limited by perceptual transparency.

Detecting and Defeating SQL Injection Attacks

Abstract —The increasing dependence on web applications have made them a natural target for attackers. Among these attacks SQL Injection Attacks (SQLIA) are the most prevalent. In this paper we propose a SQL injection vulnerability scanner that is light-weight, fast and has a low false positive rate. These scanners prove as a practical tool to discover the vulnerabilities in a web application as well as to test the efficiency of counter attack mechanisms. In the latter part of our work we propose a security mechanism to counter SQL Injection Attacks. Our security methodology is based on the design of a filter for the HTTP request send by clients or users and look for attack signatures. The proposed filter is generic in the sense that it can be used with any web application. Finally we test our proposed security mechanism using the vulnerability scanner developed by us as well as other well known scanners. The proposed security mechanism is able to counter all the vulnerabilities that were previously reported before the deployment of our security framework.

Sketch drawing by NAO humanoid robot

This paper demonstrates the sketch drawing capability of NAO humanoid robot. Two redundant degrees of freedom elbow yaw (RElbowYaw) and wrist yaw (RWristYaw) of the right hand have been sacrificed because of their less contribution in drawing. The Denavit-Hartenberg (DH) parameters of the system has been defined in order to measure the working envelop of the right hand as well as to achieve the inverse kinematic solution. A linear transformation has been used to transform the image points with respect to real world coordinate system and novel 4 point calibration technique has been proposed to calibrate the real world coordinate system with respect to NAO end effector.

Visual perception-based criminal identification: a query-based approach

Abstract The visual perception of eyewitness plays a vital role in criminal identification scenario. It helps law enforcement authorities in searching particular criminal from their previous record. It has been reported that searching a criminal record manually requires too much time to get the accurate result. We have proposed a query-based approach which minimises the computational cost along with the reduction of search space. A symbolic database has been created to perform a stringent analysis on 150 public (Bollywood celebrities and Indian cricketers) and 90 local faces (our data-set). An expert knowledge has been captured to encapsulate every criminal’s anatomical and facial attributes in the form of symbolic representation. A fast query-based searching strategy has been implemented using dynamic decision tree data structure which allows four levels of decomposition to fetch respective criminal records. Two types of case studies - viewed and forensic sketches have been considered to evaluate the strength of our proposed approach. We have derived 1200 views of the entire population by taking into consideration 80 participants as eyewitness. The system demonstrates an accuracy level of 98.6% for test case I and 97.8% for test case II. It has also been reported that experimental results reduce the search space up to 30 most relevant records.

Face liveness detection through face structure analysis

Liveness detection is a way to detect whether the person is live or not during submission of his/her biometric trait. It is mandatory in order to prevent face spoofing attacks. Therefore, in this paper, we proposed a robust face structure analysis mechanism to detect the liveness by exploiting face shape information. 3D structure/shape of the face is measured on the basis of disparity map between left and right image taken by stereoscopic vision. A gradient–based eight neighbour feature extraction technique has been proposed to extract unique features from these disparity images. It produces minimal computational cost by taking subset of the overall image. We have applied linear discriminant analysis (LDA), C–means algorithms on these features while principal component analysis (PCA) is applied on raw disparity images. We have achieved a recognition rate of 91.6%, 97.5% and 98.3% using PCA, LDA and C–means respectively, which strengthened the confidence of our proposed feature extraction technique.

Expression invariant fragmented face recognition

Fragmented face recognition suggests a new way to recognize human faces with most discriminative facial components such as: Eyes, Nose and Mouth. An experimental study has been performed on 360 different subjects which confirms that more than 80% features of the full face lies within these fragmented components. The framework intends to process each component independently in order to find its corresponding match score. Final score is obtained by calculating weighted majority voting (WMV) of each component matched score. Three different feature extraction techniques like Eigenfaces, Fisher-faces and Scale Invariant Feature Transform (SIFT) are applied on full faces and fragmented face database (ORL Dataset). It has been observed from the classification accuracy that the strength of local features (SIFT) leads to achieve an encouraging recognition rate for fragmented components whereas the global features (Eigenfaces, Fisherfaces) increases misclassification error rate. This selection of optimal subset of face minimizes the comparison time and it also retains the correct classification rate irrespective of changing in facial expression. A standard Japanese Female facial expression dataset (JAFFE) has been used to investigate the major impact on Fragmented feature components. we have obtained a promising classification accuracy of 98.7% with this proposed technique.