The phase-field fracture method is a powerful tool, as it naturally captures both crack initiation and propagation without the need to explicitly track crack paths. In phase-field fracture modeling, the decomposition of elastic strain energy plays a critical role, strongly influencing the resulting crack patterns. Various classical decomposition methods exist. For example, Miehe et al. proposed splitting the strain energy into tensile and compressive components, this strategy then widely adopted in tensile cracking scenarios. However, such decompositions fail to accurately capture the anti-crack phenomenon. An anti-crack is a counterintuitive fracture mode in which cracks propagate under compressive loading along paths similar to those observed under tension, namely extending horizontally along both sides of a central notch. This behavior has been observed in highly porous materials, where weak layer can undergo volumetric collapse under compression, creating space for crack growth. To account for the heterogeneity of porous materials and compressive scenario, instead of fully resolving the microstructure spatially, we model the material’s mechanical properties as a random field with prescribed probability distributions and spatial correlations. And the compressive strain energy is also treated as a driving force for cracks in the phase-field equations. This approach allows the porous material’s randomness to be captured. The resulting crack patterns exhibit pronounced anti-crack behavior. Additionally, we explore the potential of using the phase-field method to simulate the formation of compaction band. Prior work by our colleagues, Shegufta et al., who employed a peridynamic approach to study anti-crack behavior in porous materials, provides a valuable reference for comparison and validation of our simulation results.