False Sense of Security: Why Probing-Based Malicious Input Detection Fails to Generalize

Large Language Models LLMs can comply with harmful instructions, raising serious safety concerns despite their impressive capabilities. Recent work has leveraged probing-based approaches to study the separability of malicious and benign inputs in LLMs' internal representations, and researchers ha...

OS Fingerprints Detected

Using a combination of remote probes TCP/IP, SMB, HTTP, NTP, SNMP, etc, it was possible to gather one or more fingerprints from the remote system. While the highest-confidence result was reported in plugin 11936, “OS Identification”, the complete set of fingerprints detected are reported here...