Revisiting TESLA in the quantum random oracle model

We study a scheme of Bai and Galbraith (CT-RSA’14), also known as TESLA. TESLA was thought to have a tight security reduction from the learning with errors problem (LWE) in the random oracle model (ROM). Moreover, a variant using chameleon hash functions was lifted to the quantum random oracle model (QROM). However, both reductions were later found to be flawed and hence it remained unresolved until now whether TESLA can be proven to be tightly secure in the (Q)ROM. In the present paper we provide an entirely new, tight security reduction for TESLA from LWE in the QROM (and thus in the ROM). Our security reduction involves the adaptive re-programming of a quantum oracle. Furthermore, we propose parameter sets targeting 128 bits of security against both classical and quantum adversaries and compare TESLA’s performance with state-of-the-art signature schemes