Impact of Syntactical and Statistical Pattern Recognition on Prognostic Reasoning

This chapter provides insight on pattern recognition by illustrating various approaches and frameworks which aid in the prognostic reasoning facilitated by feature selection and feature extraction. The chapter focuses on analyzing syntactical and statistical approaches of pattern recognition. Typically, a large set of features have an impact on the performance of the predictive model. Hence, there is a need to eliminate redundant and noisy pieces of data before developing any predictive model. The selection of features is independent of any machine learning algorithms. The content-rich information obtained after the elimination of noisy patterns such as stop words and missing values is then used for further prediction. The refinement and extraction of relevant features yields in performance enhancements of future prediction and analysis.

Locally-Adaptive Naïve Bayes Framework Design via Density-Based Clustering for Large Scale Datasets

In this chapter, local conditional probabilities of a query point are used in classification rather than consulting a generalized framework containing a conditional probability. In the proposed locally adaptive naïve Bayes (LANB) learning style, a certain amount of local instances, which are close the test point, construct an adaptive probability estimation. In the empirical studies of over the 53 benchmark UCI datasets, more accurate classification performance has been obtained. A total 8.2% increase in classification accuracy has been gained with LANB when compared to the conventional naïve Bayes model. The presented LANB method has outperformed according to the statistical paired t-test comparisons: 31 wins, 14 ties, and 8 losses of all UCI sets.

Security Issues in Fog Computing and ML-Based Solutions

IoT requires data processing, which is provided by the cloud and fog computing. Fog computing shifts centralized data processing from the cloud data center to the edge, thereby supporting faster response due to reduced communication latencies. Its distributed architecture raises security and privacy issues; some are inherited from the cloud, IoT, and network whereas others are unique. Securing fog computing is equally important as securing cloud computing and IoT infrastructure. Security solutions used for cloud computing and IoT are similar but are not directly applicable in fog scenarios. Machine learning techniques are useful in security such as anomaly detection, intrusion detection, etc. So, to provide a systematic study, the chapter will cover fog computing architecture, parallel technologies, security requirements attacks, and security solutions with a special focus on machine learning techniques.

Intelligent Vision-Based Systems for Public Safety and Protection via Machine Learning Techniques

Abnormal behavior detection from on-line/off-line videos is an emerging field in the area of computer vision. This plays a vital role in video surveillance-based applications to provide safety for humans at public places such as traffic signals, shopping malls, railway stations, etc. Surveillance cameras are meant to act as digital eyes (i.e., watching over activities at public places) and provide security. There are a number of cameras deployed at various public places to provide video surveillance, but in reality, they are used only after some incident has happened. Moreover, a human watch is needed in order to detect the person/cause of the incident. This makes surveillance cameras passive. Thus, there is a huge demand to develop an intelligent video surveillance system that can detect the abnormality/incident dynamically and accordingly raise an alarm to the nearest police stations or hospitals as per requirement. If AI-supported CCTV systems are deployed at commercial and traffic areas, then we can easily detect the incidents/crimes, and they can be traced in minimal time.

Vulnerability Assessment and Malware Analysis of Android Apps Using Machine Learning

Mobile devices and their use are rapidly growing to the zenith in the market. Android devices are the most popular and handy when it comes to the mobile devices. With the rapid increase in the use of Android phones, more applications are available for users. Through these alluring multi-functional applications, cyber criminals are stealing personal information and tracking the activities of users. This chapter presents a two-way approach for finding malicious Android packages (APKs) by using different Android applications through static and dynamic analysis. Three cases are considered depending upon the severity level of APK, permission-based protection level, and dynamic analysis of APK for creating the dataset for further analysis. Subsequently, supervised machine learning techniques such as naive Bayes multinomial text, REPtree, voted perceptron, and SGD text are applied to the dataset to classify the selected APKs as malicious, benign, or suspicious.

Distracted Driver Detection System Using Deep Learning Technique

Improvement of public safety and reducing accidents are the intelligent system's critical goals for detecting drivers' fatigue and distracted behavior during the driving project. The essential factors in accidents are driver fatigue and monotony, especially on rural roads. Such distracted behavior of the driver reduces their thinking ability for that particular instant. Because of this loss in decision-making ability, they lose control of their vehicle. Studies tell that usually the driver gets tired after an hour of driving. Driver fatigue and drowsiness happens much more in the afternoon, early hours, after eating lunch, and at midnight. These losses of consciousness could also be because of drinking alcohol, drug addiction, etc. The distracted driver detection system proposed in this chapter takes a multi-faceted approach by monitoring driver actions and fatigue levels. The proposed activity monitor achieves an accuracy of 86.3%. The fatigue monitor has been developed and tuned to work well in real-life scenarios.

Cosine Transformed Chaos Function and Block Scrambling-Based Image Encryption

The authors propose an image encryption process based on chaos that uses block scrambling to reduce the correlation among the neighboring pixels and random order substitution for slightly changing the value of the pixel. The chaotic sequence for encrypting the image is generated by using two 3D logistic maps called enhanced logistic map and intertwining logistic map; the cos function helps in reducing linearity. The entire encryption process is composed of scrambling, image rotation, and random order substitution. Scrambling is used for permuting the pixels in the image so that we can reduce the correlation among the neighboring pixels, and this is followed by image rotation which can ensure that shuffling of pixels is done to the remaining pixels in the image, and at last the authors use random order substitution where they bring the small change in the pixel value. The proposed method has the capability of encrypting digital colored images into cipher form with high security, which allows only authorized ones who hold the correct secret key to decrypt the images back to original form.

Challenges and Issues in Plant Disease Detection Using Deep Learning

Deep learning (DL) has rapidly become an essential tool for image classification tasks. This technique is now being deployed to the tasks of classifying and detecting plant diseases. The encouraging results achieved with this methodology hide many problems that are rarely addressed in related experiments. This study examines the main factors influencing the efficiency of deep neural networks for plant disease detection. The challenges discussed in the study are based on the literature as well as experiments conducted using an image database, which contains approximately 1,296 leaf images of the beans crop. A pre-trained convolutional neural network, EfficientNet B0, is used for training and testing purposes. This study gives and emphasizes on factors and challenges that may potentially affect the use of DL techniques to detect and classify plant diseases. Some solutions are also suggested that may overcome these problems.

Comparative Analysis of Various Soft Computing Technique-Based Automatic Licence Plate Recognition Systems

Traffic on Indian roads is growing day by day leading to accidents. The intelligent transport system is the solution to resolve the traffic problem on roads. One of the components of the intelligent transportation system is the monitoring of traffic by the automatic licence plate recognition system. In this chapter, a automatic licence plate recognition systems based on soft computing techniques is presented. Images of Indian vehicle licence plates are used as the dataset. Firstly, the licence plate region is extracted from the captured image, and thereafter, the characters are segmented. Then features are extracted from the segmented characters which are used for the recognition purpose. Furthermore, artificial neural network, support vector machine, and convolutional neural network are used and compared for the automatic licence plate recognition. The future scope is the hybrid technique solution to the problem.

Intrusion Detection Systems

The community is moving towards the cloud, and its security is important. An old vulnerability known by the attacker can be easily exploited. Security issues and intruders can be identified by the IDS (intrusion detection systems). Some of the solutions consist of network firewall, anti-malware. Malicious entities and fake traffic are detected through packet sniffing. This chapter surveys different approaches for IDS, compares them, and presents a comparative analysis based on their merits and demerits. The authors aim to present an exhaustive survey of current trends in IDS research along with some future challenges that are likely to be explored. They also discuss the implementation details of IDS with parameters used to evaluate their performance.