Abstract Drug resistant epilepsy (DRE) accounts for 30 − 40% of all diagnosed cases of epilepsy. To date, surgical disruption of the epileptogenic zone (EZ) is the most effective treatment for seizure control in DRE. The EZ refers to a pathological brain network that is necessary and sufficient for seizures to emerge. Conventional markers localize EZ inconsistently and incompletely, which leads to only 30 − 80% of patients achieving long-term seizure freedom after the surgery. Epileptic seizures are catastrophic events initiated by several abnormal neuronal firing patterns and hyper-synchrony from EZ and might spread to other brain areas. We hypothesize that in between seizures, the EZ operates in a dynamical state that mechanistically predisposes it to having seizures. This state leads to aberrant dynamics in spontaneous activity, which can be identified with electrophysiological recordings. To test this idea,we built a generative model for cortical oscillations. When the model was controlled by strong positive feedback, it produced highly bistable oscillations in a critical-like regime of scale-free dynamics. In this regime, we observed anomalies included inhibition-dominance and hyper-irritability, therein a minuscule increase in connectivity can lead to abrupt hypersynchrony onset. Based on these modeling results, we hypothesized that using interictal data in conjunction with the ictal data likely improve EZ-localization. We retrospectively analyzed the stereo-EEG (SEEG) data from 64 focal DRE patients. Clinical experts identified the ”EpiNet” where seizure activity emerged and propagated. We assessed local dynamics and network synchrony using inter-ictal resting-state SEEG from these patient and then used the assessments to train supervised classifiers to identify EpiNet from regions that did not show seizures (NonEpiNet). Combining local and synchrony assessments yielded optimal classification on cohort and individual level. To complement these results, we next conducted unsupervised classification and identified a cohort-level tentative pathological cluster, in which EpiNet and a sizable number of NonEpiNet showed elevated 2 − 5.4 Hz synchrony with concurring excessive bistability and inhibition-dominance in 45 − 225 Hz oscillations. Combining all features yielded better EpiNet-classification than using single features (area under the curve reaching 0.85 vs 0.6 − 0.7, respectively). This explained why previous studies reported inconsistent EZ-localization with a single marker and also indicated that the aberrant dynamics of EZ indeed have a local and a large-scale aspect. The contacts forming the pathological cluster globally engaged in elevated synchrony and locally showed striking resemblance to our model in the increased seizure risk state. The NonEpiNet from this cluster did not engage in seizures and thus conventionally would have been regarded as healthy. The EpiNet contacts partitioned into the tentative healthy clusters, however, did not bear significant features of aberrant dynamics, and we postulated that these atypical EpiNet contacts could be non-essential to the EZ network required further investigation. Our findings thus offer novel evidence to support the multi-component hypothesis for EZ, and more future efforts should be directed into investigating the pathological-like NonEpiNet revealed here and with individual specificity and pathological substrates considered. Highlights Brain criticality motivated, novel biomarkers were potent to localize epileptogenic pathology Supervised and unsupervised classification were combined to characterize EZ Novel evidence was advanced to support the multi-component hypothesis of EZ Significant statement The Catastrophe theory and Brain Criticality motivated novel physio-markers were tested and found to be effective for classifying the EpiNet with the area under the receiver operating curve reaching 0.85. These novel markers, when combined with phase synchrony features, offered more accurate EpiNet-classification, indicating aberrant local and large-scale physiology collectively reflect the ”wholeness” of the EZ pathophysiology. Our results serve a strong proof-of-concept to apply this novel approach to pilot novel clinical utility with both invasive and non-invasive electrophysiological approaches.