In briefMixture-of-experts models for crypto order execution: stability, tail risk, and failure modes
SummaryThis paper compares vanilla Double Deep Q-Learning, mixture-of-experts variants, and parameter-matched dense networks for cryptocurrency order execution, using 5-minute BTC/USDT limit order book data from Binance. No learned configuration significantly beats vanilla DDQL on mean implementation shortfall, and all underperform TWAP and immediate liquidation. The authors find that annealed exploration alone eliminates policy collapse, suggesting expert partitioning masks a training-specification problem rather than providing an intrinsic benefit.
In briefAsymptotic Analysis of Trading Fees in CFMM: Studies how much trading fees liquidity providers earn from arbitrage when fee rates approach zero.
SummaryThis paper examines trading fees in automated market makers (AMMs), the dominant mechanism in decentralized finance. It derives a closed-form formula for the fees liquidity providers can receive from arbitrage trading as the fee rate approaches zero. The authors find that, when prices are continuous, arbitrage-generated fees can fully offset LVR loss. The analysis extends to price processes with jumps, concluding that jumps, along with market risk, are the only causes of LP losses.
In briefUpgrading FROST2 Threshold Signatures from TS-SUF-2 to TS-SUF-4 Security
SummaryThis paper studies FROST2, a practical threshold Schnorr signature scheme used to secure digital assets. Prior analysis showed FROST2 reaches only TS-SUF-2 security under static corruption and centralized key generation, due to TS-UF-3 attacks. The authors propose FROST2+, which adds pre-processing token verification and achieves TS-SUF-4 security under the same assumptions, also when initialized with PedPoP. Benchmarks with ZCash's FROST library report performance comparable to FROST2 and about 64-79% faster than FROST with precomputation.
In briefRateless lattice codes for secure V2X broadcast under fading and erasures
SummaryThe paper proposes a vehicular communication framework combining rateless Raptor lattice codes, nested-lattice physical-layer security with spread spectrum, and blockchain-based trust management for cooperative relaying. Under block-Rayleigh fading with random erasures, rateless Raptor reception achieves 0.3% outage with about 2.3 packets on average, outperforming LDLC and LDPC under identical channels. An eavesdropper with 8 dB combined SNR fails in 62% of sessions.
Pegah Sharifi, Hassan Khodaiemehr, Khadijeh Bagheri et al.
In briefStablecoins in the Quantum Era: A Systematization of Knowledge
SummaryThis paper is a systematization of knowledge on post-quantum stablecoins. It builds a stablecoin-specific threat model, maps cryptographic dependencies to monetary control points, and classifies migration choices by who can authorize change, where it must occur, and whether it is hybrid, post-quantum native, or a classical-component encapsulation. The authors highlight an authority-liability gap, where the party able to migrate often differs from those who bear losses. The paper also reviews interoperability proposals and identifies open problems.
Panagiotis Chatzigiannis, Navid Alamati, Suvradip Chakraborty et al.
In briefDuraShare: A BIP39-compatible secret sharing scheme with full manual recovery
SummaryDuraShare splits a BIP39 word sequence used by Bitcoin wallets among n Sharekeepers, so any k of them can reconstruct it, while k-1 colluding holders learn nothing. It combines Shamir secret sharing over F_2053, row and column checksums that detect transcription and physical errors, and an affine inner-product MAC that detects tampering. The design ensures recovery is possible using only paper, a calculator, and instructions.
In briefRANDAO Manipulation When MEV Is Present: how Ethereum's randomness can be gamed for profit
SummaryEthereum's RANDAO randomness beacon can be manipulated, and earlier work calculated the share of blocks a strategic proposer can propose. This paper introduces a generalized reward model covering forms such as consensus reward rollover, multi-block MEV, and CEX-DEX arbitrage, and gives a method for computing an ε-optimal strategy for any such reward scheme. The authors find that partial reward rollover or super-linear scaling with consecutive blocks amplifies the incentive to manipulate. They also study tail-slot slashing, modeled as a reward function, and show that honest equilibria can be recovered.
In briefA hybrid malware detector pairing autoencoders with few-shot meta-learning
SummaryThis paper proposes a hybrid deep learning framework for detecting malware, especially ransomware, when few labeled samples are available. An autoencoder compresses inputs into compact latent features, and a Model-Agnostic Meta-Learning (MAML) classifier adapts quickly to new threats from limited data. Experiments on the Ransomware Dataset 2024 across one- to fifty-shot settings report consistently high accuracy, F1 score, and Matthews Correlation Coefficient, including under extreme data scarcity.
Emmanuela Andam, Yasir Abbas Zaidi, Abdelali Hadir et al.
In briefFrom Network Intrusion Detection to Blockchain-Backed Endpoint Detection: A Survey of Decentralized Response Architectures
SummaryThis Systematization of Knowledge paper reviews research from 2019 to 2026 on blockchain-assisted intrusion detection and prevention for IoT and industrial IoT networks. It proposes a three-axis taxonomy covering detection-system class, blockchain's functional role, and response-automation maturity. The authors find that a true per-endpoint, blockchain-anchored response loop is nearly nonexistent. They also discuss cross-cutting challenges such as consensus latency, post-quantum vulnerability, and smart-contract attack surfaces, and propose a research agenda centered on hybrid on-chain and off-chain orchestration.
