Code Intention Hiding Based on AI Uninterpretability

Explore the innovative Deep Puzzling framework for concealing attack intentions and protecting code in this Hack In The Box Security Conference talk. Delve into the potential of AI algorithms in complex feature modeling, code generation, and error correction. Learn how this framework adapts to current operating environments to generate dynamic payloads, blurring the line between AI and cryptography. Discover the key aspects of data collection, feature extraction, and binary code generation modeling. Understand how AI models ensure executable binary code and the challenges of reverse analysis. Gain insights into encoding payloads within AI model parameters and the resulting complexity that hinders detection. Examine the evolution of malware, target positioning, and key security guarantees. Consider the implications of AI-powered cybersecurity and the generative, memory, and blackbox characteristics of AI in this context. Follow along with a demonstration and analysis of the framework's application, and grasp the potential impact on network security and defense capabilities.

Intro
Outline
The Evolution History of Malware
Target Positioning
Key Guaranteed Security
Debugging will Reveal Everything • Debuggabity: The adversary can discover the key role of the key
From Attack to Defend
Motivation
The boom of Al technology
Al-Powered Cybersecurity
Success Factors
The Generative Characteristic of AI
The Memory Characteristic of AI
The Blackbox Characteristic of Al
What do these Characteristics mean
The Core Idea
Method Overview
Feature Extraction
Binary Code Generation
Binary Code Repair
Main Steps
Input and Output
Model
Demonstration Target 1
Analysis
Take away