Channel state information feedback is essential for efficient communication between user equipment and base stations in wireless networks. Deep learning-based channel state information feedback models have shown great potential in compressing and reconstructing channel state information matrix, significantly reducing communication overhead. However, these models often assume ideal transmission conditions and overlook the challenges posed by interference, which can affect the reliability of feedback and reconstruction. To address these issues, we propose a Bayesian Neural Network based framework for reliable and interference-aware channel state information feedback. In specific, the proposed framework not only compresses and reconstructs channel state information but also incorporates uncertainty estimation to detect potential interference during transmission. Experimental results demonstrate that the proposed framework ensures high-quality channel state information reconstruction while promptly detecting and responding to interference, thereby enhancing the reliability of the feedback.