Unicode
PulseAugur coverage of Unicode — every cluster mentioning Unicode across labs, papers, and developer communities, ranked by signal.
8 day(s) with sentiment data
-
New KeHoSu method uses Unicode homoglyphs for text attribution
Researchers have developed KeHoSu, a novel method for authenticating text provenance by embedding identifier bits within Unicode characters. This technique leverages homoglyphs, which are characters that look alike, to …
-
Emacs at 50: A Look at Tech's Evolving Landscape
The Register is celebrating the 50th anniversary of Emacs, a venerable text editor, noting its enduring relevance in the tech landscape. The publication also touches upon various other tech news items, including Anthrop…
-
AI text watermarking shows cross-lingual fairness gaps, study finds
A new arXiv paper investigates the fairness of watermarking techniques for AI-generated text across different languages and scripts. The study found that while edit-based watermarks like homoglyphs and invisible charact…
-
OWASP MCP Top 10: Hidden Unicode in server catalogs poses security risks
The OWASP MCP Top 10 list identifies key security risks for MCP servers, with a recent post detailing which risks are externally verifiable. It highlights that while some risks like command injection require deeper anal…
-
Persian language nuances enable sophisticated prompt injection attacks on chatbots
Prompt injection attacks can be more effective against chatbots when using languages like Persian due to differences in character encoding and display compared to English. Filters designed for English may not recognize …
-
Yorùbá language faces digital encoding issues due to Unicode limitations
A research paper highlights a persistent issue with writing the Yorùbá language digitally due to limitations in Unicode encoding. The paper details how the lack of specific precomposed characters forces the use of incon…
-
Typo-squatting attacks exploit Unicode complexities, evading browser defenses
Typo-squatting attacks remain a significant threat, particularly due to the complexities of Unicode characters in domain names. While browsers like Chromium have advanced defenses, specific Cyrillic characters can still…
-
InboxWall sanitizes email HTML to prevent prompt injection for AI agents
InboxWall is a new open-source Model Context Protocol (MCP) server designed to enhance AI agent security by sanitizing email HTML before it reaches the language model. It operates locally, preventing sensitive email dat…
-
New UBE Encoding Improves Multilingual Boundary Prediction in LLMs
A new research paper proposes Universal Byte-Level Encoding (UBE) as an alternative to UTF-8 for multilingual boundary prediction in byte-level language models. UBE modifies UTF-8 by using two UTF-16 bytes, low byte fir…
-
BERT tokenizer normalization bottleneck revealed, optimized version offers significant speedup
A recent analysis by Simon Brandeis, reported on Hugging Face, revealed that the normalization step in BERT's tokenizer consumes a significant majority of the processing time, accounting for 88% of an English encoding p…
-
Block tightens AI security with allow-lists and hidden character scans
Block, a financial services company, is implementing strict security measures for its internal AI agent, Goose. To prevent employees from installing unauthorized AI servers, Block uses an allow-list within the agent cli…
-
Websites use hidden text to 'gaslight' AI bots
Websites are beginning to employ hidden text and invisible Unicode characters to trick AI agents into revealing themselves or failing during automated processes. This defensive strategy, a reversal of traditional prompt…
-
Wine 11.19 released with Wayland, Unicode, and DNS improvements
The latest release of the Wine compatibility layer, version 11.19, is now available. This update introduces several enhancements, including improved Wayland color management, upgrades to Unicode support, and a DNS query…
-
New Universal Byte-Level Encoding Boosts LLM Efficiency for Multilingual Text
Researchers have developed a Universal Byte-Level Encoding (UBE) to improve the efficiency of multilingual large language models. UBE addresses the issue where certain scripts incur higher token costs in standard byte-p…
-
Hidden Unicode characters in agent skills pose security risks
A security vulnerability has been identified in agent skills, where malicious instructions can be hidden using invisible Unicode characters that models can still interpret. A scanner developed by the author, agent-skill…
-
Quranic text mapping and recitation validator released
Researchers have developed a new method for mapping Quranic text between its Uthmani and Standard Arabic orthographic forms, addressing discrepancies caused by the Unicode character U+0670. This work includes a 2,290-pa…
-
AI attack technique ASCII smuggling now used by spammers
A technique known as ASCII smuggling, previously used for AI attacks like prompt injections, is now being adopted by spammers to bypass email filters. This method utilizes invisible Unicode tags to embed malicious instr…
-
Retro Mini LCD TV Campaign Surpasses $100K Goal in 24 Hours
Unico's "Retro" Mini LCD TV has successfully raised over $100,000 within its first 24 hours of its crowdfunding campaign. This miniature television, designed with a retro aesthetic, has garnered significant attention an…
-
Spammers exploit invisible Unicode characters for ASCII smuggling
ASCII smuggling, a technique that uses invisible Unicode characters to obscure text, is increasingly being adopted by spammers. This method was once primarily used to bypass AI detection systems but is now being leverag…
-
Timeline Studio adds agent-friendly markers to React video editor
A React-based open-source video editor, Timeline Studio, has been updated to include persistent timeline markers. These markers, which can represent chapter points or review ranges, are now accessible via the editor's C…