Search

Expectations for artificial intelligence are increasing day by day with the evolution of machine learning techniques. In recent years, machine learning has shown an increased ability to build security applications. Exceptionally high accuracy is achieved when machine learning is used for data-driven problems such as classification of malware, intrusion detection, securing health prediction systems, autonomous security audits from system logs, etc. Key challenges in adopting machine learning algorithms for security and safety critical areas include lack of transparency, inadequate vulnerability analysis, etc. Many machine learning algorithms for security and privacy are developed with little or no understanding of the underlying security aspects. Malicious external influences can cause the machine learning algorithm to predict falsely or misclassify, which might create a huge impact on security and life critical applications. The efficiency of machine learning algorithms often relies on huge volumes of data, which involves sensitive information as well. It is a greater challenge not to expose the privacy of sensitive data. This chapter addresses the impact of machine learning algorithms for applications involving medical image security.

Quantum computers are capable of processing exponentially large volumes of data in polynomial time. Quantum algorithms have created its impact in a wide range of areas, starting from simulating quantum physical systems to mathematics, cryptography information and language theory etc. Quantum image processing is a subcategory that concentrates on converting the traditional image processing algorithms for quantum computing environments. This chapter gives a brief introduction to the basics of quantum computing. It mainly focusses on quantum image representation and quantum preprocessing algorithms. Flexible representation provides a normalized state that is used for capturing the probability amplitude of the qubit containing the color information and their respective positions. The pixel intensity is stored using the basis state of the qubit by the novel enhanced quantum representation. This chapter also focusses on the Quantum image processing algorithms like filtering, edge detection, and scrambling. Matlab implementation of the quantum algorithms are also provided.

Objective: The autonomous driving allures the attention of people for its technical ability to free up the time behind the wheel and from paying attention to the road. This idea encourages the development of a variety of more sophisticated cars day by day. The driverless car is a vehicle that is accomplished by identifying its environment and steering without a car driver. In short, autonomous vehicles are designed to travel between the destinations by keeping the vehicles on a lane even when the traffic conditions change. Methods: This paper focuses on establishing a technology to develop a unique approach for the integration of autonomous robots into transportation systems. Findings: In this intended method, the autonomous vehicles consist of three subsystems. First, the input unit consists of a camera and ultrasonic sensor to sense their environment and distance measurement. Second, a processing unit is capable of analyzing the sensory data to differentiate objects on the road, which is very useful in planning a path. And the third one is a control unit that interprets the information and makes appropriate decision for keeping the vehicle in a desired path. Applications: The proposed prototype is developed using Raspberry Pi as input and processing unit. A low cost Arduino board is used as a control unit.

Objectives: In the recent past, vehicle theft is increasingly prevalent. The objective of the study is to propose a method to overcome the problem. Methods: In this paper, a technique is proposed that resolves the above mentioned problem. The primary purpose of the proposed scheme is to protect the vehicle from any unauthorized access using a cryptographic algorithm. Findings: The surety of the algorithm is achieved by combining cryptographic algorithm with a secret key. A randomly chosen binary template is used as a secret key which is then integrated with the input in order to generate the hash value using an MD5 hashing algorithm. This hash value is then compared with the stored hash value in order to operate the vehicle. In hashing algorithm, the hidden key is a data in which it is applied as an additional input to a one-way function that hashes a password. The principal use of the private key is to defend against dictionary attack and a rainbow table attack. Improvements/Applications: Thus, this technique achieves data confidentiality, data integrity, data privacy, authentication, and also is efficient in terms of processing time.

Objectives: Retrieving images from huge volumes of image database has its application in broad areas like medicine, agriculture, military etc. Annotation based approaches have become obsolete because they are time consuming and cannot describe the image effectively. The rich content in the images can overcome the limitations of annotation based techniques. Texture is the most vital visual cue used to analyze images. Methods: In the proposed technique, the image texture features are statistically represented using Generalized Gaussian Distribution in the wavelet domain. A linear score based Support Vector Machine is incorporated to identify analogous patterns to the query image from the database. Findings: The efficacy of the proposed algorithm is ascertained by conducting extensive experiments. Two texture image database of size 1400 and 1920 is used for our experiment. The proposed algorithm is verified in terms of average recall performance against the standard benchmark algorithms. It is observed that the proposed score based SVM yields higher precision and flexibility in separating the similarity within the classes and dissimilarity across different classes. Improvements/Applications: Compared to the traditional approaches, the retrieval rate of this method is improved by 30% at a considerably low computational complexity.

Objective: Monitoring the road traffic is helpful to know about the traffic density of a particular location, which helps in redirecting the traffic when crowded. In this paper, traffic density is monitored and vehicle count is estimated using background subtraction algorithm. Methods: The proposed technique utilizes a novel Motion Detection and Running Gaussian Average (MDRGA) recursive technique. The frame is captured from the input video sequence and is transformed to 0 to 255 gray levels from RGB color mode. This conversion considers the RGB values for each pixel and yield the corresponding reflectance value proportional to the brightness percentage of that pixel. The edge detection is then performed on the gray converted by locating the pixels of the image that correspond to the edges. Findings: The proposed system detects motion by comparing sequential frames crossing an imaginary line marked in the reference frame. Thus, the traffic density is detected and the proposed method is found to be efficient in terms of cost, quality and accuracy. This algorithm fetches the input video quickly from camera or storage device, processes immediately and gives the count of vehicles. This proposed system avoids occlusion problems and gives better results. Applications/ Improvements: This algorithm is helpful to avoid traffic collisions at peak areas, to avoid accidents and to reach the destination on time. Thus, the proposed system finds its application in real time systems and provides risk free environment.

Objective: In this contemporary social networking era, it is no bewilderment that, information security breach is at its epitome, especially on texts/images. Thus there is a requisite for integrity for information to be transmitted over an insecure channel. Thus image/text encryption plays a vital preprocessing role in contemporary data transmission. Method: This paper analyses the performances of four novel, lossless methodologies of implementing the enhanced Playfair cipher in spatial and frequency domains, with integer wavelet transform (IWT) through lifting scheme. Findings: The proposed methods involve a spatial domain algorithm being applied to the frequency domain of the images for encryption which gives out better desired error metrics. Application: The most feasible method out of these proposed approaches can be chosen either as a stand-alone encryption and transmission mechanism or a pre-requisite step for that specific image in steganography and watermarking.

Automatic detection of disjoint regions in synthetic aperture radar (SAR) images for applications requiring localisation and identification of objects is complicated by the nature of the speckle. A multi-level ratio of exponential weighted averages (MROEWA) together with a fast algorithm for optimal threshold selection is proposed for SAR segmentation. A non-parametric and unsupervised principle using the grey level histogram is utilised for producing the regions that are homogenous. An optimal threshold is automatically selected by maximising the separability of the classes in grey level by incorporating a simple search strategy. Experimental results on both synthetic and real images verify the effectiveness of the proposed method. It is validated that the proposed method outperforms greatly by reducing the number of arithmetic operations required for computing the optimal threshold.

The web of things, which is widely called the Internet of Things (IoT), aids as a part of our day-to-day life activities in today’s world. It has great potential to change the modern world. Although it benefits us in several ways, IoT possesses several issues, such as privacy, data confidentiality, and effective storage. The IoT community is extremely concerned about the quantum of information that may be leaked out. Hence, a secured atmosphere is mandatory for IoT devices to transmit data over public networks. In this article, a secure and optimized real-time system is proposed as an alternative methodology for data transmission from healthcare sensors to the cloud, either through local area or wide area networks. An effective optimization technique is also implemented to provide better storage.

This paper presents a medical image integrity verification algorithm which will be able to detect any changes made in the pixel value or the size of a medical image. At the same time, it also provides a secure way of transmitting images over the public domain. This algorithm can not only ensure integrity of the medical image, but also checks if the sender of the image is authentic, making it useful for archiving medical images and remote medical diagnosis. This algorithm essentially has three modules: hashing, data embedding and image encryption. Initially, hash signatures are extracted from the image and are embedded inside the medical image. In the embedding phase, the image is divided into blocks of uneven size. In the raster scanning order, the blocks are embedded with multiple pixels depending on its smoothness. At the receiver end, the signatures are extracted and it can be used to verify the integrity of the medical images. The effectiveness of the embedding algorithm is verified in terms of peak signal to noise ratio, structural similarity index metrics, correlation coefficient. L2 norm is used for verifying the integrity of the images.

Content-based image retrieval (CBIR), also known as query by image content is one of the applications of computer vision. In CBIR, the contents derived from the image like color, shapes, and textures are analyzed rather than the metadata such as keywords, tags, and/or descriptions associated with the image. In this paper, the texture features of the image are modelled using Generalized Gaussian Distribution (GGD) and Ripplet Transform. Ripplet transform has the capability of capturing structural information along with the curves compared to the traditional discrete wavelet transform. The similarity between query image and the training database is found using Kullback-Leibler Distance between GGDs. The proposed system provides greater accuracy and flexibility in capturing texture information. Experimental results on a large image database demonstrate the efficiency and effectiveness of the proposed CBIR system in the image retrieval paradigm.

Secure Content Based Image Retrieval (CBIR) deals with retrieving similar images from an image database without revealing the contents of the query image. In the traditional CBIR, it is not practically possible to store large volumes of image database locally. Hence, image database has to be stored in a remote server. In this paper, the images are encrypted using double key RSA algorithm and are stored in the remote servers. The pyramid histogram of oriented gradients (PHoG) descriptor is used as feature for representing the images. The PHoG descriptor is computed on the original image and are encrypted and then stored in the remote server. The double key RSA ensures security and the attacker cannot effortlessly decrypt the images stored in the public server.

Edge detection is a fundamental issue in automatic target detection using synthetic aperture radar (SAR) images. Edges are associated with intensity changes in the image and are efficient descriptors of the image structure. Due to the presence of speckle, edge detection in SAR images is extremely difficult. Several detectors have been developed for the detection of isolated step edges in speckled images like MRoA and RGoA edge detectors which use predefined thresholds. The modified RGoA detector defines an automatic threshold determining method. But all these edge detectors, apart from detecting the target edges, detect a number of false edges. In this paper we have proposed a new ROEWA based algorithm that automatically discriminates the object boundaries and the false edges. The principle of entropy is introduced in this classification process. Real SAR images are used to verify our method and the results are compared with the modified RGoA and entropy based MRGoA edge detectors. Experimental results show that the proposed method is robust and efficient.

The industry standard Joint Photographers Expert Group (JPEG) compression can lead to noticeable `blocking artifacts' in many cases. High frequency details of the coded images are mainly contaminated by quantization noise. The deblocking approach of JPEG compressed images using overcomplete wavelet representation can be improved by preserving the image details and reducing the effect of quantization noise as much as possible. An estimation of the actual pixel is obtained from the decoded image. The edge information is extracted from this estimated image by exploiting the cross-scale correlations among the wavelet coefficients and is protected, while the blocking effects in the smooth background regions are smoothed out in the wavelet domain without much blurring. A better improvement in visual quality is achieved if Dual Tree Complex Wavelet filters are used instead of the Dyadic Discrete Wavelet Transform.

Despite their widespread popularity, the Joint Photographers Expert Group (JPEG) compression standard can lead to noticeable 'blocking artifacts'. High frequency details of the coded images are mainly contaminated by quantization noise. The deblocking approach of JPEG compressed images using overcomplete wavelet representation can be improved by preserving the image details and reducing the effect of quantization noise as much as possible. An estimation of the actual pixel is obtained from the decoded image. The edge information is extracted from this estimated image by exploiting the cross-scale correlations among the wavelet coefficients and is protected, while the blocking effects in the smooth background regions are smoothed out in the wavelet domain without much blurring. A better improvement in visual quality is achieved if Dual Tree Complex Wavelet filters are used instead of the Dyadic Discrete Wavelet Transform.

The popular international standard for compression, the Joint Photographers Expert Group (JPEG) quite often leads to noticeable discontinuities along the boundaries called the 'blocking artifacts'. When the image is JPEG compressed the high frequency details of the coded image are mainly contaminated by quantization noise. An estimation of the actual pixel is obtained from the decoded image. The estimated image is denoised using dual tree complex wavelet transform (DTCWT), which is a variant of discrete wavelet transform. This transformation ensures approximate shift invariance and directional selectivity properties that are lacking in the traditional wavelet transform.

A new fuzzy based multiscale morphological segmentation is proposed in this paper. The technique works satisfactorily on gray scale images containing bright and dark features of various scales even in an impulse corrupted environment. The segmentation algorithm involves three passes. In the first pass, the image is preprocessed by using an iterative adaptive switching median filter which reduces the impact of impulse that causes over segmentation. In the second pass the multiple scales of bright and dark features of different objects are extracted by the respective opening and closing of the preprocessed image. The resultant image is binarized using an optimum threshold, obtained by the fuzzy Gaussian measure. The process is repeated for multiple scales of the structuring element until all the features are extracted. In the last pass, valid segments of the bright top-hat and dark bottom-hat images are detected and the contours of these images are combined to give the final segmented image. The scheme is implemented on a set of test images and the performance of the algorithm is proved better both objectively and subjectively than the standard methods. The problems of over segmentation and under segmentation are also addressed by the proposed segmentation technique.