Time-lock puzzle (TLP) is a cryptographic primitive designed to ensure that a secret remains inaccessible until a predetermined amount of time, even for an attacker with significant computational resources. These puzzles play a crucial role in various applications, from enhancing security in decentralized systems to enabling fair protocols in blockchain technology. The Chained time-lock puzzle (C-TLP) is the extension of TLP where multiple time-lock puzzles are sequentially linked, enforcing a series of time delays before a final secret is unlocked. In 2025, Kumar and Padhye proposed a C-TLP based on the integer factorization problem (IFP). They have used the design of Abadi and Kiayias and proposed a C-TLP with smaller puzzle size. In this article, we show that the scheme of Kumar and Padhye is not semantically secure and propose an improved solution, which has semantic security as well as smaller key and puzzle sizes in comparison to the C-TLP of Abadi and Kiayias.
This paper proposes a notion of branching bisimilarity for non-deterministic probabilistic processes. In order to characterise the corresponding notion of rooted branching probabilistic bisimilarity, an equational theory is proposed for a basic, recursion-free process language with non-deterministic as well as probabilistic choice. The proof of completeness of the axiomatisation builds on the completeness of strong probabilistic bisimilarity on the one hand and on the notion of a concrete process, i.e. a process that does not display (partially) inert τ-moves, on the other hand. The approach is first presented for the non-deterministic fragment of the calculus and next generalised to incorporate probabilistic choice, too.
Time-lock puzzles (TLPs) enable the transmission of messages “into the future” by requiring a specific duration to elapse before the message becomes accessible. They have a large spectrum of applications, e.g., e-voting, sealed-bid auctions, and fair multi-party computation. However, the existing time-lock puzzle schemes have some drawbacks, such as high communication overhead, excessive computing resources, or weaker security.In order to overcome these disadvantages, we provide an efficient TLP scheme under a strong security model, in which each puzzle only contains a single element in an RSA group. Our TLP scheme can resist adversaries with access to both the puzzle-generation oracle and the puzzle-solving oracle, achieving both semantic security and implicit authentication. Moreover, we introduce a new variant of TLPs called “key-recoverable time-lock puzzles” (KR-TLPs) and give a concrete construction, which offers the “solve one, get many for free” property, that is, anyone who knows the puzzle-generation key has the ability to solve puzzles in batch after time T. Furthermore, we provide two protocols for verifiability for both schemes, ensuring operational reliability while eliminating redundant computational overhead in the puzzle-solving process.
Random sketching plays a crucial role in randomized linear algebra. Although traditional dense sketching methods, such as Gaussian or Rademacher sketching, exhibit excellent stability, their computational cost remains high, specifically requiring O(nk) time to sketch an n-dimensional vector into a k-dimensional space. To address this limitation, this paper proposes an efficient and stable dense sketching algorithm. The proposed sketching matrix is implicitly generated and can embed an n-dimensional vector into a k-dimensional space with only O(n+k2) time and space complexity. For k≤O(n), this translates to a linear O(n) complexity. Owing to its implicit generation and dense structure, the algorithm is referred to as implicit dense sketching (IDS). Beyond its computational efficiency, our theoretical analysis establishes the pairwise independence of the matrix elements, norm preservation in expectation, and variance bounds, as well as guarantees for oblivious subspace embedding, low-rank approximation, and approximate matrix multiplication. To demonstrate its practical utility, we apply IDS to randomized singular value decomposition (RSVD), single-pass RSVD for streaming data, and the randomized Arnoldi process based on sketched ordinary least-squares solutions. Numerical experiments across all three applications consistently demonstrate the superior performance of the IDS algorithm over its state-of-the-art counterparts.
Real-time automata are a widely-used class of real-time systems. In this paper, two versions of strong detectability are formulated for a labeled real-time automaton (LRTA) which imply after some delay (the number of observed labels or real-time delay), along every generated infinite run, one can determine the current and subsequent states by observing the generated timed label sequence. By using the concurrent composition defined and computed in one of the author’s previous papers, necessary and sufficient conditions for the negations of the two strong versions of detectability are given. It is also proven that the verification problems for the two definitions of strong detectability are both coNP-complete. As a byproduct, we show an NP state estimation algorithm for LRTAs.
