Conscious awareness refers to information processing in the brain that is accompanied by subjective, reportable experiences. Current models of conscious access propose that sufficiently strong sensory stimuli ignite a global network of regions allowing further processing. The immense number of possible experiences indicates that activity associated with conscious awareness must be highly differentiated. However, information must also be integrated to account for the unitary nature of consciousness. We present a computational model that identifies conscious access with self-sustained percolation in an anatomical network. We show that the amount of integrated information (Φ) is maximal at the critical threshold. To the extent that self-sustained percolation relates to conscious access, the model supports a link between information integration and conscious access. We also identify a posterior "hotspot" of regions presenting high levels of information sharing. Finally, competitive activity spreading qualitatively describes the results of paradigms such as backward masking and binocular rivalry.