Deep Unsupervised Learning for Network Anomaly Detection

随着现代计算机网络的规模和复杂性的不断增加以及它们产生的大量数据，网络入侵检测是一个复杂且不断发展的挑战。在当今的网络中处理如此大量的信息超出了人类的能力，需要自动过滤机制来识别潜在的入侵，以便人类操作员进行调查。为了解决这个问题，Rank，Inc.将与多伦多大学合作开发新型异常检测系统-过滤机制，该机制可以摄取网络中的正常活动，然后随着时间的推移监控可疑威胁。这些新颖的方法基于机器学习原理，不仅可以提高检测率，还可以为人类操作员提供对已识别异常的解释，以协助他们的决策过程。开发监控网络流量的解决方案，以检测可能的网络入侵。拟议的项目旨在改进Rank，Inc.开发的工具包。通过使用先进的深度学习概念推进其异常检测套件。所提出的创新对Rank，Inc.因为他们在网络安全领域与其他解决方案提供商竞争，并将开发出独特的解决方案，使公司从直销中受益。

Network intrusion detection is a complex and evolving challenge with the increasing size and complexity of modern computernetworks and the vast quantity of data they generate. Processing this vast amount of information in today's networks is beyondhuman capabilities and necessitates automated filtering mechanisms that identify potential intrusions for human operatorinvestigation. To address this issue, Rank, Inc. will partner with the University of Toronto to develop novel anomaly detectionsystems - filtering mechanism that ingest normal activity in the network and then monitor for suspicious threats over time. Thesenovel methods are based on machine learning principles that will not only provide improved detection rate, but also supply humanoperators with explanations for identified anomalies in order to assist their decision-making process.Rank, Inc. develops solutions for monitoring network traffic to detect possible network intrusions. The proposed project seeks toimprove the toolkit developed by Rank, Inc. by advancing their anomaly detection suite with advanced concepts of deep learning.The proposed innovations are crucial for Rank, Inc. as they compete with other solution providers in the cybersecurity space andwill enable a development of a unique solution which will benefit the company from direct sales.