Search papers, labs, and topics across Lattice.
This study investigates the impact of transient errors during server-side homomorphic multiplication in the unoptimized CKKS scheme of fully homomorphic encryption (FHE). By analyzing how the timing and location of errors in ciphertext components \(c_0\) and \(c_1\) affect the correctness of the final output, the authors reveal critical vulnerabilities in FHE computations. The findings highlight that specific error patterns can lead to significant deviations in results, underscoring the need for robust error management in homomorphic encryption systems.
Transient errors in CKKS multiplication can drastically alter FHE outputs, revealing vulnerabilities that could compromise data integrity.
Homomorphic Encryption (HE) is a privacy-preserving encryption paradigm that enables computation directly on encrypted data without requiring decryption. In this paper, we study errors in fully homomorphic encryption (FHE) computations, with a particular focus on server-side homomorphic multiplication in the unoptimized CKKS (Cheon--Kim--Kim--Song) scheme. We show that both the timing and the location of errors in the ciphertext components \(c_0\) and \(c_1\) have a significant impact on the correctness of the final FHE output.
