in the Age of Artificial Intelligence and Machine Learning Cyber Defense in the Age of Artificial Intelligence and Machine Learning for Financial Fraud Detection Application

░ ABSTRACT- Cyber security comes with a combination of various security policies, AI techniques, network technologies that work together to protect various computing resources like computing networks, intelligent programs, and sensitive data from attacks. Nowadays, the shift to digital freedom had led to opened many new challenges for financial services. Cybercriminals have found the ability to leverage e- currency exchanges and other financial transactions to perform their fraudulent activities. The unregulated channel makes it essential for banks and financial institutions to deploy advanced AI & ML (DL) techniques to fight cybercrime. This can be implemented by deploying AI & ML (DL) techniques. Customers are experiencing an increase in the fraud-hit rate in financial banking operations. It is difficult to defend against dynamic cyber-attacks using conventional non-dynamic algorithms. Therefore, AI with machine learning techniques has been set up with cyber security to build intelligent models for malware categorization & intelligently sensing the fraught with danger. This paper introduces the cyber security defense mechanism by using artificial intelligence (AI), machine learning (ML)) techniques with the current Feedzai security model to identifying fraudulent banking transaction. We have given a preface to the popular ML & AI model with random forest algorithm and Feedzai’s Open ML fraud detection software tool, which provides automatic fraud-recognition to the current intelligent framework for solving Financial Fraud Detection.


░ 1. INTRODUCTION
Today in our day-to-day life, we all are living in a digital age that is interconnected to the digital ecosystems. Cyber-attacks [7] are increasing tremendously and targeting to the digital ecosystems. Cyber security has forced great impact on various critical transactions.

Cyber Security
It is the process of protecting data from various malicious attacks by using different technologies for networking servers, mobile devices, and computing systems.

Fig. 1: Cyber Security Sub-Domains
In this paper, we have given various AI and ML approaches to cyber security [13] and introduced a popular ML&AI model using random forest algorithm. Fraud detection mechanisms have been implemented using Feedzai's Open ML fraud detection software tool.
To discover new cyber-attack and to achieve higher fraud detection accuracy existing cyber fraud detection systems are not that much efficient and have faced many difficulties to detect new cyber-attack patterns.

Artificial Intelligence (AI) in Cyber Security
AI is a rapid intensification division for computer science researchers to develop new techniques, and system applications. Using AI techniques, intelligent machines can be design. AI has broad range of application in the areas like manufacturing, agriculture, grid designs, Autonomous Vehicles, Smart Cities. These smart applications are implemented using NLP [6], Chat bots & Speech Recognition, Virtual Assistants, facial recognition, and robotics. AI techniques [8] have been used in the field of cyber security for vulnerability management, breach risk prediction, incident response, exposure of threat, malware monitoring and intrusion detection and prevention etc. Artificial Intelligence techniques are considered to be as a potential solution to the increasing cybercrimes. AI can prevent and detect many abnormalities related to fraud detections.

Machine Learning (ML) in cyber security
Machine learning technology plays a vital role to address various issues of business related large-scale data. Figure 2 describes various development phases of Fraud detection using a Machine Learning model. ML approaches [1] in cyber security uses past fraud data patterns and recognizes them in their future transactions. ML algorithm like Random forest [10] (combinations of decision trees) construct decision trees to classify the data objects. It helps to find fraud traits efficiently than humans.

Fig 2: Ml Process at Cyber Security
ML is a sub-part of AI. ML approach comes with 3 core category: unsupervised learning, supervised learning, and reinforcement learning. Malware detection and network intrusion detection can be processed more efficiently using ML approaches.

Use of Deep learning (DL)
Deep Learning [2] is sub-branch of ML, which uses Neural Networks (similar to neurons in human being) techniques to simulate human brain like behavior. DL approaches behaves like human being neurons and construct the neural structural design with multifaceted interconnections [11]. DL can be subdivided across different types so-called as artificial neural networks (also known as neural nets). The neural nets are critically layered and have names as a CNN (convolutional neural network) typically used for vision (sight / pixel) processing or an RNN (a recurrent neural network) that has time based functionality [5].

░ 2. ARTIFICIAL INTELLIGENCE: A NEW TREND OF CYBER SECURITY
AI understand potential vulnerabilities and acts as an accelerator to analyze large scale of business related data and distinguish real from apparent threats.

Applications of Artificial Intelligence to Cyber security
The main goal of the artificial intelligence in cyber security is to detect cyber threats, fraud transits and to reduce the cyberattacks [14].
AI may often be better and more effective than humans in detecting malicious malware. Automation in Security improves the organization's ability to prevent and detect the damage the security flaws.

Deep Learning (Sub field of Machine learning): A New trend of Cyber Security
Detecting and preventing organization data from known and unknown cyber security [15] threats is made possible in realtime with the help of deep learning neural network algorithms.

Supervised Deep learning Model
SL model is a powerful tool to classify data and processing of data is done through machine language. In supervised learning mechanism, we use classified labeled data set. All the input information is labeled as good or bad.

Unsupervised Deep Learning Model
It is used to detect anomalous behavior in the cases of small transaction data. This approach incessantly processes the data, analyzes the new data, and updates it based on the new findings. It notices the occurrences of new data patterns and finds whether they are parts of valid or fraudulent operations. Deep learning approach in fraud detection is connected with unsupervised learning algorithms.

Reinforcement Deep learning Model
A reinforcement-learning model allows the machines to detect the ideal behavior within a specified context. It frequently learns from the environment, finds the appropriate actions to minimize the risks factor, and maximizes the rewards. A reinforcement feedback signal is used to learn its behavior.

AI & Machine learning approach for fraud detection in Banking
Many financial firms have discarded the use of legacy tools and shifted to the new-age AI & Machine learning solutions for fraud detection [12]. ML algorithms are used to practice millions of data objects quickly and link instances from unrelated datasets to detect suspicious patterns.

Machine Learning process
It consists of two main phases: Preparation Phase: Provide input data, the system is trained by labeled data. In training phase labeled data is taken and provided to the system as input. Based on this training data model is developed which is used to generate the expected output for real world input .

Supervised method for fraud detection in Banking Sectors using Random Forest Algorithm -Feedzai -AI based Open ML model
Banking Fraud Detection requires [9] a lot of effort as it contains a high risk and impact on reputation. Customer analysis is one of the biggest problems in banking sectors for analyzing the loan defaults or for detecting any fraud transaction, so keeping in mind of all these aspects we used Feedzai model.

Random forest Algorithm
In banking sector random forest algorithm is used mostly for the identification of financial frauds. Random Forest algorithm for training the dataset.

Feedzai's Anomaly Detection
We have considered Feedzai Software tool for banking fraud anomaly detections. Feedzai is one of the main intelligent platforms to solve financial crime. It is a powerful AI based anomaly detection scheme, which recognizes and stops the banking fraud transaction. Fraud transactions are assessed utilizing AI models to distinguish designs that are not clear to the natural eye.

Procedure:
Step 1: First, take the client unique exchange data and study the client profile and exchanges practices.
Step 2: In the following stage, outer information focuses are added and Feedzai's information affiliation further improves the data.
Step 3: By using machine learning models studies the transactions risk factor and detects financial crime patterns. By using the random forest model risk or fraud transactions is detected Furthermore governs like either exception discovery approach utilizing detachment woodland method or abnormality location utilizing the neural auto encoder can be utilized to adjust the choice to endorse, reject or survey. At the point when the outcome or a choice is clear, white-box clarification measure gives clear documentation.

Fraud Detection using Feedzai's-AI based Open ML Engine
Feedzai is an Open ML software Engine is used to build customized machine learning models for fraud detections.

Modules: Generic Python
The openml-generic-python module is the most powerful approach which contains a provider to load Python code that conforms to a simple API.

░ 4. RESULTS AND DISCUSSION
Implementation of machine learning model using Python with Feedzai open ML-AI based software system. We used Jupiter and Python for training the model.

Steps for Integrate the model into Feedzai open ML Engine
Step1: customer original transaction information taken into the Feedzai ML engine Step2: Prepare the data set In this model we have taken a data set historical transaction.csv file with some columns that are very common for the fraud detection use cases Timestamp, Amount, Entry mode, Card present, Fraud target Column (Which states of a transaction is fraudulent or not?) Step 3: Train the model  Step 5: Configure the fraud model external scoring service to integrate with the model that you just trained.
Website: www.ijeer.forexjournal.co.in Cyber Defense in the Age of Artificial Intelligence and Machine Learning Step 6: Save the workflow publish the result. The application is published and the model is now in production.
Step 7: Listing all the transactions Case manager in the workflow lists all the transactions that are being scored by the model that are just put into production.

░ 5. CONCLUSION
The paper we have examined the cyber security defense mechanism by using artificial intelligence (AI), machine (ML)) techniques with the current Feedzai security model to identifying fraudulent banking transaction. Summary of usage of AI and its related technologies (ML &DL), in cyber security is examined. On the other hand, presented applications of Artificial Intelligence related technologies (ML&DL) to cyber security, and have analyzed the benefit of applying deep learning to cyber security and finally deployed a latest financial fraud detection technique using supervised machine learning random forest algorithm and implemented the experiment based on data set historical transactions CSV file. An experimental result shows financial firms can detect fraud and identify genuine transactions in real time using feedzai's software open ML tool with greater accuracy. In future, there will be several topics for integration of cyber security and AI technologies can be building for Novel and special AI algorithms.