摘要
针对网络安全态势序列复杂多变，蕴含各种各样的演化规律，传统网络安全态势预测方法难以处理的问题，提出了一种专用的预测算法，该算法从长程相关的视角辨识态势序列蕴含的规律，依据事发迹象推断延续效应，经相似度、普遍性、对比度和缩放比加权后，合成预测序列。继而引入进化算法，依据预测效果调节相关参数，通过在线反馈式学习强化泛例的作用、弱化特例的干扰，提升预测算法的适应性。实验表明，该预测算法从超长态势序列中辨识多种类远距离相关性的能力很强，能对复杂多变的趋势保持自适应，预测结果更为精准可信。
Based on the view that the traditional methods for prediction of network security situation are unable to deal with the complex and inconstant network security situation sequence and its various evolution rules, the paper presents a special prediction algorithm. The algorithm identifies the rules in the situation sequence from the perspective of long range correlation, infers the subsequent effect according to occurred indication, and synthesizes the prediction sequence with the weighting by the indicators of similarity, universality, contrast ratio and scaling ratio. Afterwards, an evolution algorithm is introduced to adjust related parameters according to the prediction effect, strengthen the significance of universal cases and weaken the interference of special ones via online feedback learning, and improve the adaptability of the prediction algorithm. The experimenal results show that the prediction algorithm can perform excellently in identifying various long distance correlations from the super long situation sequence, keep self-adaptive towards complex and inconstant tendencies, and is more accurate.
出处
《高技术通讯》
CAS
CSCD
北大核心
2012年第2期140-146,共7页
Chinese High Technology Letters
基金
863计划（2007AA01Z473）和国家242信息安全计划（2007817）资助项目.
关键词
网络安全
安全态势
趋势预测
场景平移
长程相关
network security, security situation, tendency prediction, scene shift, long range correlation
