July 2025 Summaries

Server-side request forgery (SSRF) is a significant security vulnerability that allows attackers to manipulate servers into making unauthorized requests, potentially accessing sensitive internal data. This issue is particularly critical for identity infrastructure, where even minor vulnerabilities can lead to extensive data breaches, such as the Capital One incident. SSRF vulnerabilities often exploit features like user-provided URLs, overlooked DNS security, and inadequate network segmentation, making it a hidden yet potent threat. Protecting against SSRF requires multiple layers of defense, including strict validation of user inputs, network access restrictions, and secure HTTP client configurations. Stytch, a managed identity solution provider, embeds these safeguards into their infrastructure by using hardened HTTP clients, public DNS resolution, and network-layer policies to prevent unauthorized access and data leakage. These measures help mitigate SSRF risks, allowing engineering teams to focus on product development while maintaining robust security.

The Model Context Protocol (MCP) is an open standard designed to enable large language models (LLMs) to securely interact with external tools, APIs, and services by acting as an interface for AI agents to perform authorized actions. The MCP employs OAuth 2.1 for authentication, allowing users to log in and authorize AI agents through familiar web flows while maintaining data privacy and access control. This guide details the implementation of MCP authentication, focusing on OAuth 2.1's role in ensuring secure delegated access. Key components include building or integrating an authorization server, managing user consent flows, and ensuring token management and verification processes comply with MCP standards. The MCP ecosystem comprises an MCP client within the AI agent, an MCP server that translates requests into third-party API calls, and the third-party service that fulfills these requests. OAuth 2.1 updates for MCP include dynamic client registration, PKCE for enhanced security, and the mandatory use of resource indicators to ensure secure and specific token usage. By following these guidelines, MCP provides a structured, secure method for AI agents to interact with external services while maintaining user control over data access and permissions.

Stytch has introduced Trusted Auth Tokens, a feature that simplifies the integration of custom authentication methods by allowing developers to incorporate any JWT-based scheme into the Stytch environment. This tool addresses the challenge of handling nonstandard JWTs from third-party applications without the need for custom middleware by enabling developers to register an issuer profile and exchange trusted JWTs for full Stytch sessions through a single API call. Trusted Auth Tokens facilitate seamless integration with third-party platforms without requiring multiple logins, and they support a wide array of authentication scenarios, including marketplace partner integrations and auth infrastructure migrations. By removing the need for additional infrastructure, Trusted Auth Tokens empower developers to manage the complexities of authentication with ease and flexibility, ensuring compatibility and future-proofing for diverse authentication needs.

Browser fingerprinting is a technique used to identify web users based on unique browser attributes, which can aid in fraud detection and security amid rising AI-powered threats. While traditional methods like cookies can be easily spoofed, browser fingerprinting uses deterministic attributes such as user agent, IP address, and screen resolution, making it more challenging to fake. However, the effectiveness of browser fingerprinting alone is limited against sophisticated AI attacks that simulate genuine user behavior, necessitating advanced and continuously updated solutions. Device fingerprinting, a more robust method, combines browser data with device-specific signals, although it must adhere to privacy regulations and be free from exploitation. The complexity of implementing these systems leads many development teams to rely on authentication services like Stytch, which offers comprehensive fraud prevention tools, including device fingerprinting, to protect against a variety of threats while maintaining user authentication integrity. Stytch's platform analyzes billions of signals to differentiate between legitimate users and potentially harmful activities, providing businesses with the insights needed to effectively manage risk and enhance security.

System for Cross-domain Identity Management (SCIM) is an open standard that facilitates automated user provisioning and deprovisioning across multiple applications and services, reducing operational overhead for IT teams and enhancing security by preventing orphaned accounts. SCIM has become crucial for B2B SaaS companies as it enables seamless integration with existing identity infrastructures, allowing enterprises to efficiently manage employee access at scale. SCIM tools, encompassing clients like Okta and Microsoft Entra ID, service providers such as Salesforce, and third-party providers like Stytch and Auth0, offer various functionalities, including automated provisioning requests, group and role management, and compliance with industry standards like SCIM 2.0. These tools cater to different organizational sizes and needs, with options ranging from open-source libraries for custom solutions to managed services that streamline implementation and maintenance. The choice of SCIM tools depends on factors such as compliance requirements, scalability, developer experience, and pricing models, which can vary from per-connection to per-user pricing. Stytch, for example, offers a developer-friendly approach with transparent pricing and comprehensive SCIM 2.0 support, making it an attractive option for B2B SaaS companies aiming for efficient identity lifecycle management.

Stytch has launched the MCP Server, enabling developers to integrate enterprise authentication directly into their applications using AI tools like Cursor and GitHub Copilot without the need for dashboards or manual coding. This innovation, powered by Stytch Connected Apps, simplifies the setup of production auth flows by allowing AI agents to handle authentication, consent, and observability tasks. Developers can leverage OAuth-based agent authorization and Programmatic Workspace Actions to interact with Stytch assets such as projects, email templates, and redirect URLs. The MCP Server supports various MCP clients and facilitates easy deployment, such as through a command line or a simple button click for Cursor users. By streamlining the authentication setup process, Stytch aims to enhance AI-native workflows and provide a robust infrastructure for building secure, agent-ready applications.

During Liminal's Demo Day, Bobbie Chen, Stytch's Product Manager for Fraud & Security, demonstrated the company's Device Fingerprinting technology designed to combat AI-powered bot attacks that mimic human behavior in login processes. The technology analyzes various device and browser characteristics to detect and block suspicious activity in real time, thereby safeguarding users from automated threats. Chen illustrated how even simple automation scripts can be thwarted by Stytch's tools, which identify and block unauthorized actions by detecting anomalies such as headless browser automation and JavaScript property deception. Despite evolving challenges, such as increased bot activity and privacy restrictions from tech giants like Apple and Google, Stytch remains focused on fraud prevention by utilizing non-personally identifiable signals to identify potential threats. This device fingerprinting solution is available as a standalone product, allowing integration with various platforms to enhance fraud and risk prevention without relying on Stytch's authentication offerings.