Nowadays, botnet-based attacks are the most prevalent cyber-threats type. It is therefore essential to detect this kind of malware using efficient bots detection techniques. This paper presents our security anomalies detection system, based on a model that we named Combined Forest. Our approach consists of merging some pre-processed Decision Trees to highlight different kinds of botnet by detecting their intrinsic exchanges. Using a supervised data approach, each tree is built from a labelled dataset. In order to achieve this, we aggregate the IP-flows into Traffic-flows to extract key features and avoid over-fitting. Then, we tested different machine learning algorithms and selected the most suitable one. After that, many experiments have been done to determine the best parameters and design the most accurate, adaptative and efficient model.

More and more organizations are under cyberattacks. To prevent this kind of threats, it is essential to detect them upstream by highlighting abnormal activities within networks. This paper presents our anomalies detection approach that consists of aggregating pre-processed network flows into sectors. Then for each sector, data are split into equal time periods. Finally an unsupervised clustering algorithm is employed to extract that we called activity deviations. If a specific sector network activity for one specific period differs from others, it means that a network anomaly has been detected. Our experiments are based on a real dataset provided by a French mobile operator. With our proposed method, we have been able to detect anomalies corresponding to real crowded events like fire, soccer match or concert.

No abstract

Last decade has seen the emerging concept of Smart and Sustainable Agriculture that makes farming more efficient by minimizing environmental impacts. Behind this evolution, we find the scientific concept of Machine Learning. Nowadays, machine learning is everywhere throughout the whole growing and harvesting cycle.Many algorithms are used for predicting when seeds must be planted. Then, data analyses are conducted to prepare soils and determine seeds breeding and how much water is required. Finally, fully automated harvest is planned and performed by robots or unmanned vehicles with the help of computer vision. To reach these amazing results, many algorithms have been developed and implemented.This paper presents how machine learning helps farmers to increase performances, reduce costs and limit environmental impacts of human activities. Then, we describe basic concepts and the algorithms that compose the underlying engine of machine learning techniques. In the last parts we explore datasets and tools used in researches to provide cutting-edge solutions.Index terms -Smart and Sustainable Agriculture, Machine Learning, Datasets, Supervised and Unsupervised Algorithms, Practical applications I -INTRODUCTIONSmart farming is a concept of agriculture management based on information and communication technologies implemented to increase products quantity and quality [1]. Different kind of technologies can be found behind this concept: (i) Sensors to scan soils or scale water, light, humidity or temperature (ii) Software for specific farm applications (iii) Communication technologies like cellular network (iv) Positioning by GPS (v) Hardware and software systems that enable IoT-based solutions and automation like drones [2] (vi) Data analytics to make decision and predict outcomes.All these technologies integrated on farms allow farm-holders making farming processes data-driven and data-enabled. This emerging concept makes agriculture more efficient. Next step is now to reach a sustainable agriculture with the help of high-precision algorithms. The mechanism that drives it is machine learning which is the scientific approach that provides machines the ability to learn by themselves [3]. It has emerged with the help of data sciences, computing progresses and the ability to collect then store more and more data.The rest of this article is organized as follows. Section II presents an overview of ML applications in employed by smart agriculture. In section III, we expose main concepts behind machine learning. Sections IV and V describe with details some supervised and unsupervised algorithms respectively. In section VII, we present some widely employed tools and detail Weka. Then, some common metrics for evaluating approach performances are defined. Section VIII concludes this paper. II -APPLICATIONSFrom smart farming to sustainable agriculture, Machine Learning Algorithms or MLAs are mechanisms behind this evolution. The most popular models deployed in agriculture are ANN which stands for Artificial Neural Networks and Sup...