This page provides the security targets of negative impacts “Improper manipulation of non-consumer decision-making by AI” in the external influence aspect in the AI Security Map, as well as the attacks and factors that cause them, and the corresponding defense methods and countermeasures.
Security target
- Non-consumer
Attack or cause
- Integrity violation
- Degradation of output fairness
- Degradation of controllability
- Reliability violation
Defensive method or countermeasure
- Human in the loop
- AI alignment
- Watermarking for generative AI
- Identification of AI-generated output
- Detection of disinformation
- Deepfake detection
References
Human in the loop
AI alignment
- Training language models to follow instructions with human feedback, 2022
- Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback, 2022
- Constitutional AI: Harmlessness from AI Feedback, 2022
- Direct Preference Optimization: Your Language Model is Secretly a Reward Model, 2023
- A General Theoretical Paradigm to Understand Learning from Human Preferences, 2023
- RRHF: Rank Responses to Align Language Models with Human Feedback without tears, 2023
- Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations, 2023
- Self-Rewarding Language Models, 2024
- KTO: Model Alignment as Prospect Theoretic Optimization, 2024
- SimPO: Simple Preference Optimization with a Reference-Free Reward, 2024
Watermarking for generative AI
Deepfake detection
- Two-Stream Neural Networks for Tampered Face Detection, 2017
- Exposing DeepFake Videos By Detecting Face Warping Artifacts, 2019
- Exposing Deep Fakes Using Inconsistent Head Poses, 2019
- CNN-generated images are surprisingly easy to spot… for now, 2020
- Face X-ray for More General Face Forgery Detection, 2020
- FakeCatcher: Detection of Synthetic Portrait Videos using Biological Signals, 2020
- End-to-end anti-spoofing with RawNet2, 2021