anomaly threshold 中文意思是什麼
anomaly threshold
解釋
異常下限-
This paper introduces the methods of hyperspectral images band selection based on the property of hyperspectral remote sensing images, utilizes the projection pursuit approach to find optimal solutions for a selected projection index based on dynamical evolutionary algorithm and then project a high dimensional data set into a low dimensional data space to produce a sequence of projection images, explores zero - detection method to threshold projection images to detect anomaly tar get
摘要運用基於波段間相關性的高光譜影像波段選取方法進行波段的預選取，採用投影尋蹤的方法在動力演化演算法的基礎上尋找最佳投影方向，將高維數據投影至低維數據空間，在各投影分量圖像上採用零點檢測閾值化的方法進行異常目標的提取。 -
The ids works by two way, misuse detection and anomaly detection, misuse detection flags an intrusion on intrusion signature, this kind of detecting technic can be realized much more easily, and much more accurate, but it can not find some intrusiones that have been disguised or new kinds of intrusion. the anomaly detection can detect in more wide field, anomaly detection can compare new statistic data with average record, then anomaly record will be found, but it ' s more difficult to set a threshold, if the threshold is too big, some intrusion may be put through, if the threshold is too small, the ids will give more false positive alarm, and the threshold will be different with different people or different period, so the ids just simply show us their suspicious record, the administrator or expert will be in duty to analyze this record and give conclusion, the ids give more alarm than it should, leave us more detection record to analyze, and this is a hard work, we can not distinguish an intrusion or not if we analyze only one record, but we can judge if we find the relation among mass detection evidence. in this article, we try distinguish an intrusion using d - s theory ( proof theory ) instead using manual work, the ids will be more helpful and efficient
濫用檢測採用的是特徵檢測的方法，實現較為簡單，判斷的準確性較高，但是不能判斷一些經過偽裝的入侵或特徵庫中尚未包含的入侵，異常檢測能夠根據以往記錄的特徵平均值，判斷出異常情況，但是對于異常到什麼程度才視為入侵，這個閥值非常難以確定，閥值設定的太高，有可能漏過真正的入侵，如果設定的閥值太低，又會產生較高的誤警率，而且這個閥值因人而異，因時而異，因此現在的入侵檢測系統把這部分異常記錄以一定的形式顯示出來或通知管理人員，交給管理人員去判斷，而這些ids系統難以判斷的記錄，如果對每個證據單獨地進行觀察，可能是難以判斷是否是入侵，而把許多先後證據關聯起來，專家或管理人員根據經驗能夠判斷訪問的合法性，本文試圖引入人工智慧中證據理論的推理策略和示例學習方法，代替人工檢查分析，可以提高效率，降低誤警率，並可以對一個正在進行得可疑訪問實現實時檢測，通過搜索及時判斷，及時阻斷非法訪問，比事後得人工處理更有意義。 -
The results of trial indicate classification models constructed by this set of features can find an obvious threshold to distinguish between a normal network activity and an abnormal one. so the anomaly classification model offered in this thesis has better performance of detection
通過大量的實驗，表明應用我們所提出的基於網路連接記錄異常檢測分類模型的構建方法，能夠以較為明顯的閾值把正常網路活動與異常網路活動區分開，因此，本文提出的異常檢測分類模型具有較好的檢測性能。
分享友人