SolidLab
СвязатьсяContact
SolidWall AI Security Gateway

SolidWall AI Security Gateway

SolidWall AI Security Gateway is an intelligent security gateway for protecting AI applications and APIs. The solution is designed to control AI-specific interaction scenarios, protect against attacks on LLM and agentic applications, filter dangerous requests, and reduce the risk of sensitive data leakage.

SolidWall product website Download brochure

Use cases

  • Protection of AI applications against threats from OWASP LLM Top 10 Risks and OWASP Top-10 for Agentic Applications, including prompt injection, jailbreak, and sensitive data leakage.
  • Protection of web applications and AI applications from OWASP Top 10 attacks and AI-specific logical attacks.
  • Full-featured API protection with the ability to create and adjust models manually or automatically.
  • Mitigation of brute-force attacks and bots using a positive model, rate limiting, and user behavior analysis for AI-specific functionality.
  • Protection against “smart DoS” attacks that takes into account the specifics of large language model operation.
  • Control of user actions based on the analysis of legitimate transactions and business-logic control mechanisms.

Solution benefits

⌘

CPU-optimized models

We use AI models that are optimized for CPU execution and do not require expensive GPU infrastructure.

AI

Interpretable machine learning

Interpretable ML mechanisms make it possible to analyze and manually adjust resulting models, reducing the effort required to tune protection.

⊘

False-positive suppression

The solution includes quick false-positive suppression by the operator as well as fine-tuning tied to individual transaction parameters.

≋

Positive-model-based operation

The gateway uses positive application models, adaptive blocking pages, and mechanisms for detecting anomalies and significant events.

↺

Mitigation of bots and repetitive requests

The solution protects against brute-force attacks, repetitive requests, and automated abuse scenarios targeting AI-specific functionality.

▣

Monitoring and security audit

Statistics, alerts, and report exports are available, along with integration into the enterprise application delivery infrastructure.

Operating modes and fault tolerance

  • Analysis nodes can operate in active AI-application protection mode in-line between application clients and the application itself, with fast enabling or disabling of blocking mode.
  • Analysis nodes can operate in active sensitive-data protection mode in-line between an AI application and an external LLM provider, filtering requests that contain sensitive information.
  • Support for Active/Active or Active/Passive cluster configurations for analysis nodes and Active/Passive with replication for the management system.
  • Support for software bypass mode and default action execution without inspection in case of local failures or sudden load spikes.

Integrations and embedding

  • Integration with SIEM systems.
  • Integration with external systems via Syslog, SQL, and REST API.
  • Ability to load OpenAPI specifications as business-logic model actions, including as a CI/CD integration tool.
  • Integration module with the SolidPoint application security scanner for early identification of weaknesses and vulnerabilities in AI applications according to OWASP LLM Top 10 Risks.
  • API for sending data to external systems.
  • Monitoring of software and hardware component health using Zabbix and Prometheus.
  • Ability to load OpenAPI specifications as business logic model actions.