AI Guardrail Bypass and WhatsApp Privacy Exposure Highlight Security Now Roundup

Summary

Security Now examines Hidden Layer's AI guardrail bypass research, large-scale WhatsApp contact enumeration, and Microsoft's plans to add Sysmon functionality to Windows 11.

Key points

• Hidden Layer researchers reported that a carefully selected token sequence caused an older version of their prompt-injection classifier to misclassify a test prompt as benign.
• The researchers said the technique can also produce false alarms, potentially undermining trust in defensive models.
• WhatsApp researchers reported probing over 100 million phone numbers per hour without effective blocking or rate limiting, exposing account information and, in some cases, profile photos and text.
• The host said the WhatsApp researchers subsequently confirmed that Meta had implemented effective rate limiting.
• Microsoft plans to make Sysmon functionality available in Windows 11; the tool logs events useful for threat hunting and forensic investigations.
• The hosts discussed the tradeoff between pinning npm dependencies to avoid malicious updates and automatically receiving security patches.
• The discussion distinguished an EU proposal for voluntary content screening from an earlier proposal for mandatory scanning; future legislative steps remained planned.

Tools

Related Articles

• Learn Malware Analysis with Ghidra and x64dbg Through Ransomware DemosDavid BombalSysmon
• Detecting REVSTEALER Activity and Automating File Removal With Wazuhwazuh.comSysmon
• Wazuh Demonstrates Detection of Common Windows Privilege Escalation Techniqueswazuh.comSysmon
• Researcher Demonstrates OAuth Token Theft Through a Sideloaded Windows AppX Hosthuntress.comSysmon
• How to Build Cybersecurity Skills and Experience From Freshman YearCYBERWOXSysmon