Ciphertext-Policy Attribute-Based Encryption (CP-ABE) has revolutionized access control mechanisms by enabling fine-grained data sharing based on user attributes. However, conventional CP-ABE schemes are vulnerable to privacy breaches due to plaintext exposure of access policies, potentially revealing sensitive user information. In response, this paper introduces a novel approach utilizing Generative Adversarial Networks (GANs) to enhance the privacy and efficiency of CP-ABE in IoT environments, especially on resource-constrained mobile devices. GANs are employed to generate synthetic access policies that closely mimic real policies without disclosing actual user attributes. The generator in the GAN framework produces synthetic policies, while the discriminator ensures their authenticity and resemblance to real policies. By attaching these synthetic policies to ciphertexts, our method obscures sensitive attribute information, thereby mitigating privacy risks. The effectiveness of the proposed approach is evaluated through extensive simulations, impressively, our model achieved an Frechet Inception Distanc (FID) score of 1.35 after just 230 epochs of training. This score demonstrates a high level of fidelity, indicating that our approach effectively generates synthetic access policies that closely resemble real-world counterparts. This research contributes a robust framework for leveraging GANs in cryptographic protocols, offering a practical solution to enhance CP-ABE’s privacy capabilities while maintaining rigorous access control mechanisms.