This joint session includes two talks from Kennesaw State University. Machine Learning and Cybersecurity Analytics using the NSL-KDD Dataset - Zularbine Kamal, Kennesaw State University
Technology significantly impacts our everyday lives with the communication network serving as the backbone of all technological advancements. As these networks expand, they present multiple points of vulnerability that attract hackers. Consequently, every communication system must implement a robust intrusion detection mechanism to ensure security.
HPCC Systems offers a comprehensive suite for storing raw data, preprocessing, model storage, and running queries through Roxie. In this presentation, we will demonstrate how to leverage the HPCC Systems Machine Learning Library with the NSL-KDD cybersecurity dataset.
HPCC Systems currently supports several machine learning algorithms, both supervised and unsupervised, and makes them available via machine learning bundles. In this project, we have leveraged all the algorithms that have data classification capability to detect network intrusion with the dual goal of getting the highest accuracy possible for the trained model while ensuring efficiency during its training. We will also demonstrate the use of the “Myriad Interface”, which can perform multiple independent machine learning tasks within a single interface invocation. Invoking the activities in parallel allows them to be distributed across the nodes in the cluster, thereby maximizing the performance while minimizing the run time. Lastly, we will also cover how the ML_Core.Preprocessing bundle can be used for data preprocessing including label encoding, scaling, and one-hot-encoding.
The content of this presentation is aimed at individuals working with machine learning, cybersecurity specialists, and HPCC Systems users and technology enthusiasts overall.
-------------------------------------------------------------
Model Inversion Attacks with the HPCC Systems Platform - Andrew Polisetty, Kennesaw State University
In this world of machine learning, feeding the model with inputs and training the model is one thing, and securing that model is another. Since many companies leverage machine learning models for decision-making using sensitive data, attackers can target these data-sensitive models, and one of the biggest threats to these models is MIA (Model Inversion Attack). Mainly, MIA is a technique that can be leveraged to reconstruct sensitive information such as financial data. Attackers can access these models and gather predicted data, which can be used as input for training a new model similar to the original one. The attackers can then infer the sensitive data from the original model to reconstruct the training data or build a comparable model.
In this project, we have leveraged a public loan dataset and utilized the HPCC Systems platform to perform black-box attacks and to design solutions to prevent them. We first developed the machine learning model and then utilized it to build the original, threat and defender models. For the original model, we chose a credit risk assessment scenario consisting of a person’s loan and personal details. We developed the original model by using the learning trees algorithm from the HPCC Systems machine learning bundle. In this scenario, the attackers would access the inputs and outputs of the model through querying, where they can analyze the data points to perform a black-box attack. In the attack model, we simulate an attack by querying the original model and we train the attack model using its output. For the prevention model, we are currently exploring different approaches, such as adding noise to the output of the original model to manipulate the attacker. So far, the raw accuracy obtained from the learning trees algorithm is still relatively low, so we are also training models using logistic regression and continuing to explore different defenders for the prevention model.
© 2024 LexisNexis Risk Solutions
Музыка
Фильмы
ТВ Шоу
Категории видео
