Alper Çakan

Alper Çakan, Vipul Goyal, Omri Shmueli

Quantum fire is a distribution of quantum states that can be efficiently cloned, but cannot be efficiently converted into a classical string. First considered by Nehoran and Zhandry (ITCS'24) and later formalized by Bostanci, Nehoran, Zhandry (STOC'25), quantum fire has strong applications and implications in cryptography, along with important connections to physics and complexity. However, constructing and proving the security of quantum fire so far has been elusive. Nehoran and Zhandry gave a construction relative to an inefficient quantum oracle. Later, Bostanci et al gave a candidate construction based on group actions, however, even in the oracle model they could only conjecture the security of their scheme, and were not able to prove security. In this work, we give a construction of public-key quantum fire relative to a classical oracle and prove its security unconditionally. Going further, we introduce two stronger notions that generalize it: Quantum key-fire where the clonable fire states serve as keys, and interactive (i.e. LOCC) security for quantum (key-)fire. We give a construction of quantum key-fire relative to a classical oracle and unconditionally prove that it satisfies interactive security for any unlearnable functionality. As a result, we also obtain the first classical oracle separations between various notions in physics and cryptography: *** A computational separation between two fundamental principles of quantum mechanics: No-cloning and no-teleportation, which are equivalent in information-theoretically. *** A separation between copy-protection security (Aaronson, CCC'09) and LOCC leakage-resilience security (Cakan, Goyal, Liu-Zhang, Ribeiro, TCC'24). *** A separation between computational no-cloning security and no-learning security, two notions introduced recently by Fefferman, Ghosh, Sinha, Yuen (ITCS'26).

Simos Gerasimou, Hasan Ferit Eniser, Alper Sen et al.

Deep Learning (DL) systems are key enablers for engineering intelligent applications due to their ability to solve complex tasks such as image recognition and machine translation. Nevertheless, using DL systems in safety- and security-critical applications requires to provide testing evidence for their dependable operation. Recent research in this direction focuses on adapting testing criteria from traditional software engineering as a means of increasing confidence for their correct behaviour. However, they are inadequate in capturing the intrinsic properties exhibited by these systems. We bridge this gap by introducing DeepImportance, a systematic testing methodology accompanied by an Importance-Driven (IDC) test adequacy criterion for DL systems. Applying IDC enables to establish a layer-wise functional understanding of the importance of DL system components and use this information to assess the semantic diversity of a test set. Our empirical evaluation on several DL systems, across multiple DL datasets and with state-of-the-art adversarial generation techniques demonstrates the usefulness and effectiveness of DeepImportance and its ability to support the engineering of more robust DL systems.