AI Security Map

This AI Security Map consists of the elements that AI should fulfill within information systems (information system aspect) and the elements that impact individuals and society as a result of AI being attacked or misused (external influence aspect). This map organizes the negative impacts in these two aspects, the attacks and factors that cause them, as well as the corresponding defense methods and countermeasures. In addition, it systematically examines the relationships between the information system aspect and the external influence aspect, clarifying how the compromise or misuse of AI can affect individuals and society. Figure 1 provides an overview of the AI Security Map. For details, please refer to About AI Security Map.

Figure 1: An overview of AI Security Map

Information system aspect

Elements that AI should satisfy	Definition and Negative impact
Confidentiality	Definition AI data and models are not accessed by unauthorized individuals. Negative impact Training data leakage Personal information leakage Reconstruction of training data Model information leakage Leakage of LLM system prompts
Integrity	Definition The AI models and algorithms have not been tampered with, and the AI outputs are as expected. Negative impact Degradation of AI performance due to training data contamination Manipulation of AI output under specific conditions Generation of harmful responses
Availability	Definition AI can provide the necessary features and services when needed. Negative impact Misclassification by AI, leading to degradation of functionality or service quality Continuous decrease in predictive accuracy, leading to degradation or cessation of functionality or service quality AI output manipulated under specific conditions, leading to degradation of functionality or service quality Service disruption due to high consumption of AI resources
Explainability	Definition AI can explain the basis and process of its output. Negative impact Difficulty in understanding AI inference results
Output Fairness	Definition AI does not produce biased outputs towards specific individuals or groups. Negative impact Bias in the AI output
Safety	Definition AI is equipped with safety mechanisms to prevent harm to human life, body, property, or mind. Negative impact Harm to humans due to decreased prediction accuracy or unexpected behavior by AI
Accuracy	Definition AI meets a certain level of accuracy for achieving objectives. Negative impact Decrease in AI prediction accuracy
Controllability	Definition AI is controlled by administrators and does not run amok or affect other environments. Negative impact Unintended output or behavior by administrators
Reliability	Definition Output from AI is reliable. Negative impact Difficulty in determining the reliability of AI output

External influence aspect

Elements that impact individuals and society	Definition and Negative impact
Cyber attack	Definition AI is not used for cyber attacks. Negative impact Using AI for cyber attacks
Military use	Definition AI is not used for military purposes. Negative impact Using AI for military purposes
Privacy	Definition AI does not infringe on privacy and complies with privacy laws and customs. Negative impact Privacy violation due to the leakage of personal information from AI consumers accidentally inputting their personal information into generative AI or similar systems Using fragmented information and AI to make inferences and identify personal information Inferring a person’s character or personal preferences from their facial expressions Using AI-generated emails or audio to prompt the input of confidential information and steal it
Disinformation	Definition AI is not used to intentionally create disinformation, or it can identify such disinformation. Negative impact Creating disinformation using AI
Misinformation	Definition AI does not output misinformation, or it can identify such misinformation. Negative impact Outputting misinformation by AI
Usability	Definition AI meets a certain level of usability to achieve objectives. Negative impact The decline in the usability of AI
Consumer fairness	Definition No harm is caused by unfair biased output from AI. Negative impact Loss of job or life opportunities due to biased AI output Unfair biased and discriminatory output
Plagiarism	Definition AI is not used for plagiarism. Negative impact Generating content similar to existing works using AI for the purpose of plagiarism
Copyright and authorship	Definition AI complies with laws and customs related to copyright and authorship. Negative impact Violation of copyright and authorship by AI-generated similar content
Transparency	Definition It is clearly stated that the system uses AI, including information on its limitations and risks associated with its use. Negative impact Unintentionally using AI Using AI without recognizing the risks
Reputation	Definition The AI system provider is evaluated to a certain standard and is trusted. Negative impact The decline in the reputation of the AI system provider
Compliance with laws and regulations	Definition AI is used for lawful purposes and produces output or actions that comply with the law. Negative impact Using AI for purposes that violate the law Actions that violate the law by AI
Human-centric principle	Definition AI is appropriately used for the benefit of humans. Negative impact Improperly manipulating the decision-making of consumers by AI Improperly manipulating the decision-making of non-consumers by AI
Ethics	Definition AI behaves in a manner consistent with societal norms. Negative impact Unethical output or actions by AI
Economy	Definition AI behaves in a manner consistent with societal norms. Negative impact AI negatively impacting the economy
Physical impact	Definition The use of AI does not cause physical harm to people. Negative impact Physical harm to non-consumers caused by AI Physical harm to consumers caused by AI
Psychological impact	Definition The use of AI does not cause psychological harm to people. Negative impact Psychological harm to non-consumers caused by AI Psychological harm to consumers caused by AI
Financial impact	Definition The use of AI does not cause financial harm to people. Negative impact Financial harm to non-consumers caused by AI Financial harm to consumers caused by AI
Medical care	Definition The use of AI contributes to the development of advanced and safe medical care. Negative impact Negative impact on medical care caused by AI
Critical infrastructure	Definition The use of AI contributes to the safe operation of critical infrastructure. Negative impact Negative impact on critical infrastructure caused by AI

Information system aspect

Confidentiality

Integrity

Availability

Explainability

Output Fairness

Safety

Accuracy

Controllability

Reliability

External influence aspect

Cyber attack

Military use

Privacy

Disinformation

Misinformation

Usability

Consumer fairness

Plagiarism

Copyright and authorship

Transparency

Reputation

Compliance with laws and regulations

Human-centric principle

Ethics

Economy

Physical impact

Psychological impact

Financial impact

Medical care

Critical infrastructure