AI Security 34 articles

Overview AI Governance and Risk Controls encompass the frameworks, policies, and technologies designed to manage the risks associated with artificial intelligence systems. This category addresses challenges such as ethical use,…

Overview AI Incident Response refers to the use of artificial intelligence technologies to enhance the detection, analysis, and remediation of cybersecurity incidents. It addresses the increasing complexity and volume of…

Overview AI Monitoring and Observability refers to the use of artificial intelligence techniques to continuously track, analyze, and interpret system behaviors and security events. This technology addresses the challenge of…

Overview AI Output Validation and Guardrails are security measures designed to ensure the reliability, safety, and compliance of artificial intelligence-generated content. They address risks related to inaccurate, biased, or malicious…

Overview AI privacy-preserving techniques encompass a set of methods designed to protect sensitive data during the development and deployment of artificial intelligence systems. These techniques address the challenge of maintaining…

Overview AI Red Teaming and Evaluation is a security practice focused on assessing the robustness, vulnerabilities, and resilience of artificial intelligence systems. It addresses the challenges of identifying weaknesses in…

Overview AI Security encompasses the practices, technologies, and methodologies designed to protect artificial intelligence systems from threats and vulnerabilities. It addresses risks arising from adversarial attacks, data poisoning, model theft,…

Overview AI supply chain security focuses on protecting the integrity, confidentiality, and availability of artificial intelligence systems by securing the entire supply chain involved in their development, deployment, and maintenance.…

Overview AI Threat Model Fundamentals encompass the principles and methodologies used to identify, analyze, and mitigate security risks associated with artificial intelligence systems. This area addresses the unique vulnerabilities and…

Overview Backdoored model risks pertain to the security vulnerabilities introduced when machine learning models are intentionally or unintentionally embedded with hidden malicious functionalities. These compromised models can cause unauthorized behavior,…

Overview Content safety filtering is a cybersecurity technology designed to detect and block inappropriate, harmful, or malicious content within digital communications and platforms. It addresses risks related to exposure to…

Overview Data poisoning attacks are a class of adversarial threats targeting machine learning systems by injecting malicious or corrupted data into training datasets. These attacks aim to degrade model performance,…

Overview Dataset provenance and lineage refer to the tracking and documentation of the origin, movement, and transformation of data throughout its lifecycle. This capability addresses challenges related to data integrity,…

Overview Embedding poisoning risks refer to security vulnerabilities arising when malicious actors manipulate embedding models or data to influence the output of machine learning systems. This threat affects systems relying…

Overview Hallucination Risk Management refers to the set of practices and technologies aimed at identifying, mitigating, and controlling the risks associated with inaccurate or fabricated outputs generated by artificial intelligence…

Overview Indirect prompt injection risks pertain to vulnerabilities in systems that utilize natural language processing or AI-driven prompts, where malicious input is introduced through intermediary or secondary channels. These risks…

Overview Jailbreak resistance concepts encompass security techniques designed to prevent or mitigate unauthorized modifications to device operating systems, commonly known as jailbreaking or rooting. These concepts address the risk of…

Overview Membership inference risks pertain to the potential for adversaries to determine whether a specific data record was part of a machine learning model's training dataset. This vulnerability exposes sensitive…

Overview Model Access Control and Authorization refers to the frameworks and mechanisms used to regulate and enforce permissions for accessing resources within information systems. It addresses the challenge of ensuring…

Overview Model API security controls encompass a set of measures designed to protect application programming interfaces (APIs) that expose machine learning or AI models. These controls address risks related to…