Epilepsy is a chronic brain disorder that affects approximately 60 million people worldwide. Approximately 30% of people with epilepsy do not respond to treatment with one or more medications or to resective surgery. It is well known that the occurrence of epileptic seizures produces a series of sudden dynamic changes in EEG signals manifested as partial or generalized seizures in the epileptic patient. In the present study, a dissimilarity index (DI) was developed for the detection of epileptic seizures in EEG signals based on an abrupt change detection model, supported by one-class classifier obtained from support vector data description learning machine. DI was estimated using Poincaré plot features including the complex correlation measure from the EEG signals which were used such as the inputs to the one-class classifiers. DI showed that at the seizure onset its value increases during following epochs. It was clearly evident that the DI revealed a change in the statistical distribution of the sets before and after the time instant of the seizure onset. It was shown that the SVDD based dissimilarity index for epileptic seizure detection is a good parameter to characterize the epileptic seizure of the patient.