Autism is a neurodevelopmental disorder characterized by poor social abilities, communication deficiency and restricted behavioural patterns. Recently, scholars started to consider the possibility of detecting biological markers for better and faster diagnosing autism. This problem has been approached from different perspectives considering biochemical, neurophysiological, and neuroanatomical markers. Following this perspective, our intent was to investigate whether a structural brain signature of autism can be detected in children by using a whole brain morphological analysis. To this aim, we selected 43 male children with autistic spectrum disorder and 46 male controls, matched for age. Structural brain images (T1 image), intelligence scores (Full IQ, Verbal IQ, Performance IQ), and Autism Diagnostic Observation Schedule (ADOS) scores were considered for analyses. Source-Based Morphometry, a multivariate method based on Independent Component Analysis to detect maximally independent cortical networks of gray matter differences was applied to autistic and control brains. Results showed a statistically different network between ASD children and controls, including several cerebellar regions (Inferior Semi-lunar lobule, Tuble, Uvula, Pyramis, Declive, Cerebellar Tonsil) and the Fusiform Gyrus, confirming, but, also expanding previous results. In addition, separate temporal, frontal, and parietal networks were found to be significantly correlated with the Stereotyped Behaviour ADOS scores. These morphologic differences may be particularly useful in paving the way for future objective methods to diagnose autism.