CTI View: APT Threat Intelligence Analysis System
Open Access
- 3 January 2022
- journal article
- research article
- Published by Wiley in Security and Communication Networks
- Vol. 2022, 1-15
Abstract
With the development of advanced persistent threat (APT) and the increasingly severe situation of network security, the strategic defense idea with the concept of “active defense, traceability, and countermeasures” arises at the historic moment, thus cyberspace threat intelligence (CTI) has become increasingly valuable in enhancing the ability to resist cyber threats. Based on the actual demand of defending against the APT threat, we apply natural language processing to process the cyberspace threat intelligence (CTI) and design a new automation system CTI View, which is oriented to text extraction and analysis for the massive unstructured cyberspace threat intelligence (CTI) released by various security vendors. The main work of CTI View is as follows: (1) to deal with heterogeneous CTI, a text extraction framework for threat intelligence is designed based on automated test framework, text recognition technology, and text denoising technology. It effectively solves the problem of poor adaptability when crawlers are used to crawl heterogeneous CTI; (2) using regular expressions combined with blacklist and whitelist mechanism to extract the IOC and TTP information described in CTI effectively; (3) according to the actual requirements, a model based on bidirectional encoder representations from transformers (BERT) is designed to complete the entity extraction algorithm for heterogeneous threat intelligence. In this paper, the GRU layer is added to the existing BERT-BiLSTM-CRF model, and we evaluate the proposed model on the marked dataset and get better performance than the current mainstream entity extraction mode.Keywords
This publication has 17 references indexed in Scilit:
- Information Extraction of Cybersecurity Concepts: An LSTM ApproachApplied Sciences, 2019
- Cyberthreat Detection from Twitter using Deep Neural Networks2019
- YEDDA: A Lightweight Collaborative Text Span Annotation Tool2018
- Acing the IOC Game2016
- Automatic Extraction of Indicators of Compromise for Web Applications2016
- Towards a Relation Extraction Framework for Cyber-Security Concepts2015
- Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation2014
- Extracting Cybersecurity Related Linked Data from Text2013
- Extracting Information about Security Vulnerabilities from Web Text2011
- Framewise phoneme classification with bidirectional LSTM and other neural network architecturesNeural Networks, 2005