Mobile Adhoc Network (MANET) is vulnerable to network attacks due toudits open communication medium. Blackhole and wormhole attacks are the mostudsevere attacks in the network. The attacks cause congestion and increase theudpossibility of con�dential data theft. Unfortunately, the existing security solutionsudare insu�cient to protect the network. This work proposed a new securityudframework, named Extra Secure Adhoc on Demand Distance Vector (ESAODV).udThis framework provides a defense-in-depth protection through layered securityudmeasures: secure protocol and intrusion detection system (IDS) with extraudcountermeasures. The �rst layer implements lightweight packet authentication,udand the second layer monitors and counters malicious packets. In this study,udESAODV was implemented using Java in Time Simulator/Scalable WirelessudAdhoc Network Simulator, and analyzed using R-Statistics, Sigma Plot andudMinitab. Results showed that ESAODV had contained the blackhole attackudand the hybrid blackhole attack (HBHA) e�ectively. The number of corruptingudrouting tables of benign nodes could be minimized to be near zero even if theudnumber of attackers were increased. In addition, the IDS accurately detectedudthe wormhole and the variant of wormhole attack called diversion of packet overudthe wormhole link (DP-WHL). The false positive for live attack detection wasudsmall. The accuracy of detection was more than 94.5 percent. Although attackersudchanged the pattern of packets diversion, the IDS detected the new attack patternudin near real time. In addition to these �ndings, this research has also modeledudfour performance metrics data of ESAODV, i.e., memory usage, elapsed timeudfor completing routing tasks, number of route replies and route success, basedudon both linear regression and neural network. Goodness of �t parameters forudthe models based on the neural network was higher than the linear regression.udESAODV has been proven to provide a comprehensive protection from the mostudsevere attacks in the network. Furthermore, the performance metrics of ESAODVudbased on the neural network produced a superior model.
展开▼
