Frequently Asked Questions
Several organizations are continuously working in the field of AI Security. NIST, ISO, IEEE, ENISA, ETSI, and GARD could be some of the organizations to have in-depth insights into the topic. There are some open-source frameworks like MITRE ATLAS or Bosch Assessment tool that organizations could use.
With the advancement in technology, there have been increasing AI algorithm failures. The complexity and number of such losses have been growing every day. Looking at some of the facts and current industry scenario indicates the need for AI Security, and correct measures must be in place before a significant model failure.
- According to Microsoft, 89% of organizations have the right tools to secure their AI.
- Gartner found that 60% of AI providers will include a means to mitigate possible harm to their AI assets by 2024
- IBM quotes that 80% cost difference between cyberattack scenarios, where secure AI was developed vs. not deployed.
AI security refers to the domain of securing AI systems. One can also refer to cybersecurity for AI applications & models/algorithms. AI Security focuses on identifying, detecting, preventing, remediation, information, investigating and responding to malicious attacks on AI Systems. The field is concerned with AI model vulnerabilities, studying attacker capabilities, consequences of attacks, and building algorithms that can resist security challenges.
AIShield is the last layer of defense in an existing device, network & application security. It protects AI models, which are highly invested and most valuable assets of an AI system. AI model and system security is not covered under existing Cybersecurity measures since AI security attacks can happen as part of an AI application’s input, output, and data payload where the hacker poses as a legitimate user.
AIShield protects AI systems against four types of adversarial attacks.
Model Extraction – The act of stealing and extracting a proprietary model
Model Evasion – The act of making a model do the wrong thing making it intentionally biased
Data Poisoning – The act of making a model learn wrong things by injecting malicious samples
Model Inference – The act of making the model reveal its logic and data