Recent reviews on the use of experience-based food insecurity scales in the Indian context suggested the addition of "how often" related items to food insecurity modules to avoid overestimation of food insecurity, especially in underprivileged communities. Following this recommendation, we adapted the 8-item Food Insecurity Experience Scale (FIES), an official tool for measuring access to food within the Sustainable Development Goals (target 2.1), and assessed its validity and reliability in socially-backward communities in the Indian context. The polytomous Rasch model was successfully applied and soundly integrated within the probabilistic methodology already in use for the FIES, allowing the computation of comparable prevalence of food insecurity at different levels of severity and related measures of uncertainty. Data from the SWABHIMAAN programme survey, which collected information on food insecurity from mothers of children under two years of age in three Indian states (Bihar, Odisha, and Chhattisgarh), was used for analysis. Results suggest that the proposed adapted version of the FIES can be considered as a proper tool for measuring food insecurity in underprivileged communities, since it satisfies requirements of internal and external validity and reliability. Individual determinants and protective factors of food insecurity were also investigated within this methodological framework and results suggest that education, economic wealth, and homestead kitchen garden can act as a buffer against food insecurity, while the number of pregnancies seems to exacerbate a situation of food insecurity.