Daily Edition · Page 5

THE DAILY INDEX

Technology, AI, software, security, crypto and business

ATTACKS & VULNERABILITIES

TLDR InfoSec

HIDING LINUX PROCESSES WITH BIND MOUNTS

This post refines a methodology for hiding Linux processes by using a bind mount to mount a different directory on top of /proc/PID for the malicious process. The author was looking to refine the original methodology by creating a more realistic directory to overlay but realized that they could instead overlay the proc directory of a spontaneous process which is usually ignored by sysadmins. They note that this method would still be easy to detect if you are aware of the technique and provide a command to do so.

IMPLEMENTING SECURITY INVARIANTS IN AN AWS MANAGEMENT ACCOUNT

A Permission Boundary defines the maximum permissions that can be applied to a given principal. In an AWS management account, permissions boundaries are necessary because SCPs cannot be applied. This post introduces a set of security invariants that should be true in an AWS management account and provides an IAM policy to enforce them. It also links to a Lambda function that can be used to attach the policy to all newly created IAM principals.

SCHEMING REASONING EVALUATIONS

Apollo Research tested 6 popular LLMs to determine how they would react when given a prompt that gave them a goal, instructed them to achieve their goal at all costs, and then gave them a task to review files that contain proof that the developers have a different goal. 5 out of the 6 models tested would attempt to circumvent hurdles or safeguards to their goals and in some cases, models would lie about the actions on consultations. The researchers found that some models could deliberately underperform or otherwise scheme to attempt to achieve their goals.

TLDR IndexPage 5