Secure your AI workloads against AI attacks and security threats with industry-first patented technology. Track and automatically respond to any attack incidents in real-time by leveraging Splunk SIEM integration.

AIShield is an AI-security product designed to protect AI-powered devices in the face of emerging security threats. It provides automated hacker-level vulnerability analysis and end-point protection to harden AI systems against newer vulnerabilities such as model theft/extraction, data poisoning, algorithm evasion & model/data inference attacks. AIShield easily integrates with Microsoft Sentinel to deliver real-time alerts. It protects the IP and brand of organizations against critical breaches and attacks on AI systems (devices, assets, workloads, models).

Product Technical Details:

• Enterprise-class AI model security vulnerability assessment and threat-informed defense generation

Core features:

• Vulnerability scanning - Analysis of various types of AI/ML models against attacks such as theft, poisoning, evasion, and inference. Extraction and Poisoning attacks for image classification, sentiment analysis, time series forecasting/classification, and tabular classification are currently available.
• End-point protection - Threat-informed defense generation and availability of attack data for native hardening of model
• Intrusion detection prevention - Real-time prevention and monitoring of new attacks in the cloud and on devices
• Threat intelligence feed - Active threat hunting and incident report triggers
• Splunk Integration - Report security incidents via SIEM connectors to Splunk; Threat hunting capabilities aided by vulnerability analysis and active monitoring.

Usability & Support:

• Accessible: AIShield is available in cloud-native SaaS configurations developed with an API-first approach and detailed dashboards for various stakeholders across all industries.
• Flexibility: AIShield is compatible with leading AI development frameworks, toolchains, and software to enable flexibility and seamless integration. It works with encrypted AI/ML models or API end-point of AI/ML models. Direct support for TensorFlow and indirect support for other ML frameworks.
• Ease of implementation: Easy-to-use APIs with ready reference implementations in Jupyter Notebooks, product guides, POSTMAN configuration files, and API documentation. Easy integration with MLOps platform with product API. SIEM/SOAR connectivity via containerized defense (customer to deploy).

Additional attributes:

• Supports 200+ attack types across 20+ models and data type variations (e.g.: image classification, time series forecasting etc.)
• Integration and deployment of end-point defense mechanisms along with the original model in target environments such as cloud or devices
• Frequent attack database updates and threat hunting capabilities (e.g., OSINT, research, academia)
• Threat-informed defense model available in ONNX format with alert telemetry sent in OCSF compliant schema.