Deep Learning (DL) systems are key enablers for engineering intelligent applications due to their ability solve complex tasks such as image recognition and machine translation. Nevertheless, using DL in safety- security-critical requires provide testing evidence dependable operation. Recent research this direction focuses on adapting criteria from traditional software a means of increasing confidence correct behaviour. However, they inadequate capturing the intrinsic properties exhibited by...

Deep Learning (DL) systems are key enablers for engineering intelligent applications. Nevertheless, using DL in safety- and security-critical applications requires to provide testing evidence their dependable operation. We introduce DeepImportance, a systematic methodology accompanied by an Importance-Driven (IDC) test adequacy criterion systems. Applying IDC enables establish layer-wise functional understanding of the importance system components use this information assess semantic...

Is it possible to comprehensively destroy a piece of quantum information, so that nothing is left behind except the memory whether one had at point? For example, various works, most recently Morimae, Poremba, and Yamakawa (TQC 2024), show how construct signature scheme with certified deletion where user who deletes on m cannot later produce for m. However, in all existing schemes, even after still able keep irrefutable evidence was signed, thus they do not fully capture spirit deletion. In...

Quantum information allows us to build quantum money schemes, where a bank can issue banknotes in the form of authenticatable states that cannot be cloned or counterfeited. Similar paper banknotes, existing banknote consists an unclonable state and classical serial number, signed by bank. Thus, they lack one most fundamental properties cryptographers look for currency scheme: privacy. In this work, we first further develop formal definitions privacy schemes. Then, construct public-key...

Quantum no-cloning theorem gives rise to the intriguing possibility of quantum copy protection where we encode a program in state such that user possession k states cannot create + 1 working copies. Introduced by Aaronson (CCC 09) over decade ago, has proven be notoriously hard achieve. In this work, construct public-key encryption and functional schemes whose secret keys are copy-protected against unbounded collusions plain model (i.e. without any idealized oracles), assuming (post-quantum)...

Motivated in part by applications lattice-based cryptography, we initiate the study of size linear threshold (`t-out-of-n') secret-sharing where reconstruction function is restricted to coefficients {0,1}. We also complexity such schemes with additional requirement that joint distribution shares any unauthorized set parties not only independent secret, but uniformly distributed. prove upper and lower bounds on share schemes, measured total number field elements distributed parties. our...

Quantum secret sharing (QSS) allows a dealer to distribute quantum state among set of parties so that certain subsets can reconstruct the secret, while unauthorized obtain no information. While QSS was introduced over twenty years ago, previous works focused only on existence perfectly secure schemes, and share size known schemes is exponential even for access structures computed by polynomial monotone circuits. This stands in contrast classical case, where efficient computationally-secure...

Deep Learning (DL) systems are key enablers for engineering intelligent applications due to their ability solve complex tasks such as image recognition and machine translation. Nevertheless, using DL in safety- security-critical requires provide testing evidence dependable operation. Recent research this direction focuses on adapting criteria from traditional software a means of increasing confidence correct behaviour. However, they inadequate capturing the intrinsic properties exhibited by...