Conventional biomarkers, e.g., spikes and high-gamma oscillations, localize epileptogenic zones (EZ) inconsistently, which may contribute to variable postsurgical seizure freedom (30~80%) in drug resistant epilepsy (DRE) patients. The novelty of this research entails application of the criticality hypothesis (brains benefit from operating between order and disorder) to new biomarker development for improving EZ-localization. We advanced novel markers for high seizure risk brain areas and validated them in conjunction with connectivity features on simulated data and on narrow-band frequency SEEG resting-state recordings from DRE patients with focal onset (n=64, 29.7±9.5 yo, 29 females). The criticality and synchrony features were used to train supervised classifiers to localize the epileptogenic network (EpiNet) identified by physicians. For generalization of the EZ mechanisms, the narrow-band features of all SEEG samples (m=7,138) were dimension reduced to a handful number of eigen feature coefficients. Two unsupervised classifiers were subsequently employed to delineate sample clusters in the eigen feature space. Supervised classification revealed that combining criticality and connectivity features yielded better EpiNet-classification than using individual features (area under receiver operating characteristics curve reaching 0.85 vs 0.6~0.7, respectively), which explains why single markers may localize EZ inconsistently. Unsupervised classification revealed a cohort-level pathological cluster of samples that globally engaged in strong synchrony and locally showed high amplitude bistability, high inhibition, and aberrant scaling – a striking resemblance to our model in a high seizure risk phase in the critical regime. Further investigation discovered a smooth, funnel-shaped surface in a low-dimensional eigen feature space, and the individual positions on the surface were predictive of individual variability in supervised EpiNet-classification. Importantly, a particular area on the surface, samples was associated with higher probability of observing interictal spikes, wherein many nonEpiNet and EpiNet were indistinguishable by their eigen features. As the rising phase of the spikes has been shown to be predictive of incoming seizures, we suggest that this low-dimensional surface in the eigen feature space characterizes an interictal continuum. With these results, we conclude that EZ dynamics encompass anomalies at both the local and network level, and the EZ may comprise both EpiNet and pathological nonEpiNet.