Safety Alignment Can Be Not Superficial with Explicit Safety Signals

Recent studies on the safety alignment of large language models LLMs have revealed that existing approaches often operate superficially, leaving models vulnerable to various adversarial attacks. Despite their significance, these studies generally fail to offer actionable solutions beyond data...

Design/Logic Flaw

PHP before 5.1.3-RC1 might allow remote attackers to obtain portions of memory via crafted binary data sent to a script that processes user input in the htmlentitydecode function and sends the encoded results back to the client, aka a "binary safety" issue. NOTE: this issue has been referred to a...

CVE-2006-1490

CVE-2006-1490 affects PHP up to version 5.1.3-RC1. It describes a remote information disclosure where crafted binary data sent to a script processing input in html_entity_decode may leak memory contents. The issue is labeled as a memory leak (information disclosure) rather than a true memory corr...

CVE-2006-1490

PHP before 5.1.3-RC1 might allow remote attackers to obtain portions of memory via crafted binary data sent to a script that processes user input in the htmlentitydecode function and sends the encoded results back to the client, aka a "binary safety" issue. NOTE: this issue has been referred to a...