The decomposition step is a major bottleneck in the effectiveness of the index-calculus algorithm for solving the discrete logarithm problem on the Jacobian of hyperelliptic curves with low genus, as it requires solving a nonlinear system of multivariate polynomial equations. Sarkar and Singh, in the year 2016, proposed a decomposition method for the curves, defined over quadratic extensions, which avoids solving multivariate polynomial systems. In this paper, building on the Sarkar-Singh work and using the cyclic group of the underlined extension field, we identify a class of weak curves for which the decomposition is easy to handle. In the new decomposition, we completely avoid solving the system of multivariate polynomial equations, and the time complexity remains the same as that of the Sarkar-Singh decomposition. The new decomposition applies to specific curves defined over extension fields. For a given random hyperelliptic curve, it can be efficiently checked whether this decomposition attack applies to it or not, using the discrete logarithms in the cyclic group of the underlying extension field. The discrete logarithm problem (DLP) over finite fields is generally easier than the DLP over the Jacobian of hyperelliptic curves. Therefore, we can reasonably assume that discrete logarithms over the underlying finite field are accessible. This assumption aids in determining whether the given curve is susceptible to this type of attack.
The Internet of Medical Things (IoMT) has emerged as a transformative domain in healthcare, offering unprecedented opportunities for remote monitoring, smart diagnosis, and automated medical services. However, the rapid development and deployment of IoMT applications have inadvertently attracted the attention of attackers, leading to an increased risk of security breaches. Attackers will use malware to exploit IoMT systems to get unauthorized access to them. To address this challenge, the proposed work introduces a Multi-View Contrastive Graph Neural Network (MVC-GNN) for malware detection in IoMT environments. The proposed approach leverages graph-based deep learning and contrastive representation learning to effectively distinguish benign and malicious network traffic to enable timely and accurate detection of malware. This work uses CICIoMT 2024 dataset, a state-of-the-art publicly available dataset, for a comprehensive evaluation of the proposed approach. Experimental results highlight the high efficacy of the approach, with an accuracy of 99.69
Programming bugs and flaws can have fatal consequences for the security of cryptographic software and may allow an attacker to bypass authentication, forge signatures, decrypt sensitive data, or even completely reveal secret keys. Certain categories of bugs, such as subtle carry-propagation flaws in large-integer or prime-field arithmetic carried out by many public-key cryptosystems, manifest only under very specific and, therefore, extremely rare input conditions, which makes them hard to detect with conventional software testing methodologies. While there exist a few papers that describe such Hard-to-Find Bugs (HFBs) and study their security implications, a more comprehensive treatment and systematization are still lacking. The present paper aims to fill this gap and analyzes the challenges posed by HFBs in software implementations of public-key cryptosystems. More concretely, we define and categorize HFBs, provide a survey of HFBs that have been found in widely-used open-source cryptography libraries (some of which remained undetected for up to 10 years), and discuss the benefits and limitations of common testing and prevention techniques, including differential testing, static analysis, fuzzing, formal verification, and Known Answer Tests (KATs) tailored to HFBs. Raising awareness of HFBs is important for software developers and security auditors who implement and test cryptographic algorithms for mission-critical systems where correctness and robustness are paramount. By shedding light on subtle implementation flaws and how to reduce their occurrence, this paper contributes to improving the real-world security of public-key cryptosystems.
SNOW-V is a stream cipher designed by Ekdahl et al. (ToSC 2019(3)) and mainly used for protecting 5G network communications. Existing works on SNOW-V focused on optimizing its performance with respect to 16-bit registers which are part of ciphers’ linear feedback shift registers. In this paper, we present a 64-bit implementation of SNOW-V by extending all core round components and state update functions of the cipher to operate on 32 bits. We verify the correctness of our implementation with SNOW-V test vectors and provide benchmarking results on two different software platforms – Intel Core i7 and AMD Ryzen 5. Our results demonstrate a throughput gain of 90
Many keystream generator designs rely on mapping internal state bits to the output via a Boolean function. This introduces correlation between the keystream and the generator’s state, as well as a functional dependency, rendering the generator vulnerable to various cryptanalytic techniques such as correlation, algebraic, and distinguishing attacks. These attacks can often be leveraged for key recovery. A core challenge in stream cipher design, therefore, is to develop generators producing keystream with strong statistical properties while ensuring security beyond vulnerability to known cryptanalytic methods—specifically, achieving a security level where key recovery is, in principle, limited only to the exhaustive key search. In this work, we address this challenge with the Random Transposition Cipher (RTC), a generator inspired by the experiment of drawing balls with replacement from an urn. This analogy forms the basis of a hidden Markov model (HMM) of the random transposition shuffling process and constitutes the core of the RTC, superficially resembling the RC4 stream cipher. We present a detailed security analysis of the RTC, demonstrating its resistance to correlation, algebraic, and distinguishing attacks. Furthermore, we establish that key recovery for the RTC is theoretically infeasible by any method other than the exhaustive search.
