Publication: Sistema de Detección de ataques EDoS en entornos Cloud
Loading...
Official URL
Full text at PDC
Publication Date
2015
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
La computación en la nube constituye el modelo de provisión de recursos de cualquier plataforma o aplicación web como servicio, bajo demanda y de forma dinámica (elasticidad) a través de internet. Se concibe como un nuevo paradigma informático que permite ofrecer los distintos servicios a través de la red. Esto aporta un importante conjunto de ventajas sobre los servidores físicos convencionales, pero también implica nuevos desafíos. De entre ellos cabe destacar el rápido crecimiento de las brechas de seguridad, que viene propiciado por la adaptación de diferentes tipos de amenazas convencionales, siendo los ataques de Denegación de Servicio Distribuido (DDoS) una de las más frecuentes. A los ataques de denegacion de servicio que buscan agotar la sostenibilidad económica de los servicios en la nube se les denomina ataques EDoS (Economic Denial of Sustainibility). Se trata de una amenaza muy reciente, que ha puesto en alerta a la comunidad investigadora y a las diferentes organizaciones para la ciberdefensa. Sin embargo, a pesar del peligro que conllevan, existen muy pocas propuestas para detenerlos. Con el fín de contribuir a su mitigación, en este trabajo se propone una estrategia de detección de intrusiones especializada en el reconocimiento de estos tipos de ataques. La aproximación realizada combina la construcción de métricas a partir del análisis de la entropía del tráfico monitorizado, con la construcción de series temporales y modelos predictivos capaces de predecir el estado de la red. Esto permite identificar comportamientos impredecibles, y por lo tanto, anómalos, que revelen la presencia de amenazas EDoS. La propuesta ha sido evaluada con la colección de muestras de dominio público CAIDA'07, y bajo su despliegue en un caso de uso real. En ambos casos se ha comportado de manera eficaz y precisa, demostrando su capacidad de afrontar los objetivos propuestos
The cloud computing model is the provision of resources for any platform or web application as a service, on demand and dynamically (elasticity) through internet. It is conceived as a new computing paradigm that can offer different services through the network. This provides an important set of advantages over conventional physical servers, but it also involves new challenges. Among them include the rapid growth of security breaches, which is facilitated by the adaptation of different types of conventional threats, with the Distributed Denial of Service (DDoS) attacks one of the most frequent. For the denial of service attacks that seek exhaust the economic sustainability of cloud services are called EDoS attacks (Denial of Economic Sustainibility). This is a very recent threat, which has alerted the research community and various organizations for cyber defense. However, despite the danger they entail, there are very few proposals to stop them. In order to contribute to mitigation, in this paper a strategy intrusion detection specialist in the recognition of these types of attacks is proposed. The approach made, combining the construction of metrics based on the analysis of entropy of the monitored traffic, with the construction of time series and predictive models capable of predicting the state of the network. This allows to identify unpredictable behavior, and therefore anomalous, indicating the presence of EDoS threats. The proposal has been evaluated with the collection of samples CAIDA'07 public domain and under their deployment in a case of actual use. In both cases it has acted efficiently and accurately, demonstrating its ability to meet the objectives
The cloud computing model is the provision of resources for any platform or web application as a service, on demand and dynamically (elasticity) through internet. It is conceived as a new computing paradigm that can offer different services through the network. This provides an important set of advantages over conventional physical servers, but it also involves new challenges. Among them include the rapid growth of security breaches, which is facilitated by the adaptation of different types of conventional threats, with the Distributed Denial of Service (DDoS) attacks one of the most frequent. For the denial of service attacks that seek exhaust the economic sustainability of cloud services are called EDoS attacks (Denial of Economic Sustainibility). This is a very recent threat, which has alerted the research community and various organizations for cyber defense. However, despite the danger they entail, there are very few proposals to stop them. In order to contribute to mitigation, in this paper a strategy intrusion detection specialist in the recognition of these types of attacks is proposed. The approach made, combining the construction of metrics based on the analysis of entropy of the monitored traffic, with the construction of time series and predictive models capable of predicting the state of the network. This allows to identify unpredictable behavior, and therefore anomalous, indicating the presence of EDoS threats. The proposal has been evaluated with the collection of samples CAIDA'07 public domain and under their deployment in a case of actual use. In both cases it has acted efficiently and accurately, demonstrating its ability to meet the objectives
Description
Trabajo de Fin de Grado en Ingeniería Informática (Universidad Complutense, Facultad de Informática, Curso 2014-2015)