A Unified Framework for Compressed and Encrypted Text Direct Processing

The rapid growth of textual data calls for systems that can efficiently manage large-scale text, while the need to safeguard text privacy has become increasingly urgent. Homomorphic encryption enables computation directly on encrypted data but incurs significant ciphertext expansion and high storage and computation costs. Compression, by exploiting redundancy, reduces storage and transmission overhead, and homomorphic compression further allows direct processing without decompression. Combining these techniques offers a promising path to both privacy and efficiency. We propose Double Homomorphism, a novel theoretical framework that integrates compression and encryption to support direct processing of doubly-encoded textual data without the need for decoding. To demonstrate its practicality, we design DOHO, a text management system built upon the DH framework. DOHO supports direct modification, random access, and analysis over compressed and encrypted textual data. We evaluate DOHO using five real-world textual datasets. In the efficiency-prioritized setting, DOHO achieves up to 69.04× throughput improvement for text modification, 3.54× speedup for text analysis, and 3.75× enhancement in space efficiency compared to uncompressed solutions. In the privacyprioritized setting, DOHO outperforms state-of-the-art FHEbased methods, delivering $\mathbf{1 0 8. 0 5} \times$ throughput improvement for text modification, $\mathbf{9. 6 9} \times$ speedup for text analysis, and $\mathbf{1 2. 0 6} \times$ space efficiency improvement.