Réduction à l'aide d'une approche basés sur les KNN des alertes fausses positives issues d'un détecteur d'intrusions réseau

الموضوعات

علوم تطبيقية وتكنولوجية

الملخص

Nous avons proposé, au titre de cette thèse, de contribuer à apporter des éléments de réponse à la problématique de réduction des alertes fausses positives issues d'un système de détection d'intrusions en utilisant une approche basée sur la règle KNN, moyennant la conception et l'évaluation de quatre nouveaux algorithmes, fondés sur les concepts de colonies de fourmis et de mémétique. Les résultats obtenus, à la suite des tests sur le benchmark KDD'99, se sont avérés satisfaisants et démontrent que les approches proposées permettent effectivement de réduire le taux d'alertes fausses positives, avec des taux de détection importants

روابط وملفات

التعريف والنوع

رقم الوثيقة

fe6569c8-1bb8-4903-92e5-f6d55c320395

رقم العقد

0

نوع الوسائط

Crawler

نوع المحتوى

الرسائل العلمية

صيغة المصدر

رسائل دكتوراة

نوع الملف

word

أسماء الملفات

1379563_1.docx, 1379563_2.pdf

بيانات النشر

ألقاب المؤلفين

[{"name_ar":"Ahmed Riadh Baba- Ali","title_ar":"اشراف","title_en":"Supervision"},{"name_ar":"Amal Miloud Aouidat","title_ar":"اعداد","title_en":"Preparation"}]

اللغة

French

المصدر والدورية

اسم المصدر

Réduction à l'aide d'une approche basés sur les KNN des alertes fausses positives issues d'un détecteur d'intrusions réseau

المحتوى والصفحات

عدد الصفحات

0

ترجمة الملخص

We have proposed in the title of this thesis, to help solve the problem of reducing false positive alerts from a system of intrusion detection using an approach based on the K- nearest neighbors rule (KNN), through the development and evaluation of four new algorithms, based on the concepts of ant colonies and memetics, in the prospect of improving intrusion detection while reducing the rate of false alarms.Noting that a system of intrusion detection based on KNN rule uses a training set of data for classification, whose accuracy is dependent on the size of the said set, and that this size determines the processing time, we explored the applicability of the concepts respectively of ant colony and memetic through two approaches of instance selection.The first approach is an aggregation of KNN alerts classification with two new algorithms, one abbreviated Clust ANT-IS, which is based on the principles of antcolony optimization and DBS CAN clustering, the other, an instance selection algorithm,abbreviated ALAIS improves his previous and is based on active learning. These two algorithms are built on the basis of the Ant-IS algorithm, which is an instance selection algorithm, based on the principles of ant colony optimization. This algorithm improves the performance of standard 1NN, selecting instances to create effective training set.Ant-IS has been proven effective in the field of classification.The second approach involves the pairing of an improved genetic algorithmabbre-viated GIS to a new controlled local search algorithm, abbreviated CLS. This twinninghas created the MCLS algorithm. The latter, improves the capabilities of standard KNN classifier in intrusion detection. What is new in this method is the proposed process of local search that guide both mutation and selection of the solution.The results, after tests on KDD’99 benchmark , are very satisfactory and show that both approaches can effectively reduce false positive alert rates of 0.60% (ALAIS) and 0.53% (Clust Ant-IS) for the approach based on ant colonies, and 1.05% (GIS) and 0%(MCLS) for the evolutionary approaches, with respective rates of detection : 98.64%,99.38%, 98.02% and 99.87%

كلمات الباحثين

Algorithmes génétiques, Ant Colony Optimization, Condensing, False positive, Instance Selection,Set Reduction, Intrusion Detection, KNN, Local Search, Machine Learning, Memetic Algorithm, Mesure, Métaheuristiques, Network Security, Optimisation par colonies de fourmis, Réseaux d'ordinateurs, أمن الشبكات, اختيار المثال, البحث المحلي, التكليف, تحديد مجموعة, كذبة ايجابية, مستعمرات النمل

إشراف وإعداد

الإشراف

Ahmed Riadh Baba- Ali

الإعداد

Amal Miloud Aouidat

الاقتباسات الببليوغرافية