Hiding in Plain Sight: Secure Communication with Language Models
A new approach to steganography leverages the power of large language models to conceal information with provable security guarantees.
Science
Outsmarting Attackers: A Strategic Approach to Cryptographic Algorithm Selection
![The AttackerDP algorithm, when constrained by a 0.202-second execution limit, exhibits a threshold of approximately 310,310 methods-a value determined by the simulation’s parameters, including a benchmark of 500 methods with [latex]k=500[/latex] and [latex]v=1000[/latex], a success probability range of [latex]s\_{.,j} \in [0.05, 0.85][/latex], and method costs between 40 and 200 units, calculated via a linear cost function [latex]\varphi(x) = x[/latex]-suggesting that system performance is intrinsically tied to the interplay between computational budget and the defined parameter space.](https://arxiv.org/html/2604.21436v1/Images/dp_time_threshold_crossing.png)
This research introduces a game-theoretic model to optimize the combination of cryptographic algorithms, anticipating and responding to evolving attacker strategies.
Scaling Beyond Blockchain: A Deep Dive into Layer-2 Solutions
This review unravels the complex world of Layer-2 scaling technologies, explaining how they’re poised to unlock mainstream blockchain adoption.
Securing Communications: A Head-to-Head Coding Challenge
New research rigorously compares three leading methods for protecting data transmitted over noisy channels, revealing performance differences in practical, short-message scenarios.
Seeing Isn’t Believing: The Growing Threat to Self-Driving Car Sensors
A new review synthesizes the landscape of attacks targeting autonomous vehicle perception, revealing critical vulnerabilities as systems increasingly rely on multiple sensors.
Consistent Choices: Replicable Bandit Algorithms for Reliable Decision-Making
This review explores how to build multi-armed and linear bandit algorithms that consistently select the same actions under repeated runs, improving the reliability of reinforcement learning systems.
Scaling Attention: A New Approach to Long Contexts
Researchers have developed a method for significantly extending the length of sequences processed by attention mechanisms without requiring hardware modifications.
Whispers in the Wireless: Hiding Data in Wi-Fi Signals
![A Wi-Fi steganography system conceals data within the seemingly innocuous structure of wireless network traffic, offering a covert communication channel by exploiting the inherent redundancies and flexibility of the [latex]802.11[/latex] protocol.](https://arxiv.org/html/2604.20521v1/x1.png)
Researchers have developed a system to embed secret messages within the normally unused components of Wi-Fi transmissions, offering a new avenue for covert communication.
Containing the AI Frontier: A Formal Verification Approach
New research details how rigorous pre-deployment checks can bolster the security of AI sandbox infrastructure and prevent potentially catastrophic model escapes.
Beyond Sums and Differences: Homomorphic Encryption Gets a Comparative Edge
A new technique streamlines privacy-preserving computation by efficiently combining arithmetic and comparison operations within fully homomorphic encryption schemes.