Scalable and Deep Anomaly Detection from Big Data with Similarity Hashing

异常检测是一项重要的大数据分析技术，旨在识别数据挖掘中异常但有意义的模式。大数据的性质要求一种能够处理快速发展的不同类型数据的检测方法。然而，现有的检测方法要么计算成本高，要么检测性能差。该项目旨在开发一个检测框架，以提高检测性能和效率，基于一种新的深度学习模型，称为深度隔离森林，不同于传统的基于人工神经网络的模型。这一结果将为智能制造中的实时预测性维护、网络安全中的入侵检测等各种应用带来巨大的好处。

Anomaly detection, aiming to identify anomalous but insightful patterns in data mining, is an important big data analytics technique. The nature of big data requires a detection method that can handle fast-evolving data of diverse types. However, existing methods suffer from either high computational cost or low detection performance. This project aims to develop a detection framework to advance detection performance and efficiency, based on a novel deep learning model called deep isolation forest which is different from the traditional artificial neural network based models. The outcome will bring huge benefits to various applications such as real-time predictive maintenance in smart manufacturing, and intrusion detection in cybersecurity.