Yasna brings together recent headlines from selected sources and makes them easier to sort with tags, filters, and search.
Search across headline titles and summaries.
Weekly headline count for the current query.
New Research Uncovers Tokenizer Blind Spots in Leading LLMsSubtle obfuscation techniques can systematically evade the guardrails that today's large language models rely on. Researchers from Mindgard team found that adversaries can "smuggle" malicious payloads past tokenizers using emojis, zero-width spaces and homoglyphs.
In this blog entry, we uncovered a campaign that uses fake GitHub repositories to distribute SmartLoader, which is then used to deliver Lumma Stealer and other malicious payloads. The campaign leverages GitHub’s trusted reputation to evade detection, using AI-generated content to make fake repositories appear legitimate.
The Chinese APT hacking group "Mustang Panda" has been spotted abusing the Microsoft Application Virtualization Injector utility as a LOLBIN to inject malicious payloads into legitimate processes to evade detection by antivirus software. [...]
The Chinese state-sponsored threat actor known as Mustang Panda has been observed employing a novel technique to evade detection and maintain control over infected systems
Hackers are targeting Windows machines using the ZIP file concatenation technique to deliver malicious payloads in compressed archives without security solutions detecting them. [...]
The Cyber Police of Ukraine has announced the arrest of a local man who is suspected to have offered their services to LockBit and Conti ransomware groups