Design and Implementation of Machine Learning and Big Data Analytics models for Cloud Computing platforms

Abstract

There are concerns regarding the protection of significant digital assets because it is noticeable that cyber attackers are outstripping defenses. AI models need specialized cybersecurity and defense keys in order to lower dangers, enhance information privacy, and enable a secure federated knowledge situation. The growth of “artificial intelligence” has controlled the rise of several new fields, including “machine learning” (ML), “natural language processing” (NLP), CPU vision, and several more. Massive volumes of information are produced in the “Internet of Things” (IoT) age and are acquired from a variety of heterogeneous sources, including “mobile devices”, “sensors”, and “social media”. Big Data faces significant challenges in terms of processing, storage, and analytical capabilities. Firewall security has proven to be insufficient as a result of significant restrictions against external attackers. Given that computer worms and viruses, which are intelligent semi-autonomous agents, are responsible for the majority of network-centric cyberattacks, it has become necessary to combat them with intellectual semi-autonomous mediators that can identify, evaluate, and respond to cyberattacks. Specified that the majority of network-centric cyberattacks are produced by CPU worms and illnesses, which are intelligent semi-autonomous agents, it has become important to combat them with intelligent semi-autonomous agents that can recognise, assess, and respond to cyberattacks.

The study aims to determine the Big Data Analytics and Machine Learning paradigms for usage in cybersecurity and to make use of big data analytics and machine learning.

This study used a case study research methodology. This is because each statistics analytics model for cybersecurity is observed as a unique case that needs to be examined in its own context. Case studies have been used a lot in prior research on cybersecurity. The investigator knows two data analytics models or frameworks through a review of the literature and a study of the 8-person sample. There were eight people interviewed in total. Even if there might not be much data, it is sufficient for the current stages of this investigation. Future studies may look at other publications to discover other data and analytics models that are pertinent to cybersecurity.

Table 4 illustrates the overall show of our CART procedure in predicting scores on the banks on the training dataset. On the testing data, the precision of the approach was 83.1%. On the test dataset, the approach got a kappa of 88.76% and a reliability of 92.7%. On the training dataset, the SVM figure's overall accuracy for predicting bank health remained at 79.1%. The Kappa statistic and Kappa SD were 67.6% and 0.14, combined. On the testing data, the approach had an efficiency of 92.7% and a kappa of 88.59%. The efficiency of the randomized forest in the training phase was 85.57 %. Large amounts of information are is for connections and trends, leading to the creation of algorithms for these kinds of relationships and patterns. Passive sources of information include laptop data such as IP address, information security certifications, keypad usage, clickstream trends, and WAP data.