API Gateway Security Using ML Request Screening

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

APIs exposed to external applications are vulnerable to malicious attacks, necessitating secure exposure techniques.

Innovation Solution

Implementing a machine learning model to predict potentially malicious requests and prevent them from reaching the network, while allowing non-malicious requests to proceed.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If the API is exposed to external third party applications, then the API enables third party applications to access network services, but the network becomes vulnerable to malicious attacks via the API

Engineering Contradiction:

ImproveAPI accessibility to third party applicationsVSAvoidmalicious attacks on the network

Core Design Contradiction:

Adaptability or versatilityVSObject-affected harmful factors

Solution Approach 1:

A machine learning model is introduced as an intermediary component between the API and the network. This model analyzes incoming requests and predicts whether they are malicious, acting as a mediator that filters harmful traffic while allowing legitimate requests to pass through to the network services.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The machine learning model performs preliminary analysis of requests before they reach the network. By predicting malicious intent in advance and blocking suspicious requests proactively, the system prevents potential attacks rather than reacting to them after they occur.

Inventive Principle:

Principle #10Preliminary action

2Reliability

If traditional security methods are used to protect the API, then the network is protected from attacks, but the API functionality and performance are degraded

Engineering Contradiction:

Improvenetwork security protectionVSAvoidAPI request processing efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

Traditional mechanical security filtering methods are replaced with a machine learning-based predictive system. The ML model uses pattern recognition and analysis to identify malicious requests, providing more accurate and efficient security protection compared to rule-based or signature-based traditional methods.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The security approach transitions from static rule-based filtering to dynamic machine learning-based prediction. The system continuously learns from request patterns and adapts its security parameters, improving both protection effectiveness and processing efficiency by making security decisions based on learned behaviors rather than fixed rules.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20250350611A1System, method, and computer program for application programming interface (API) security

Publication Date: 2025.11.13 AMDOCS DEV LTD

AI summary

As described herein, a system, method, and computer program are provided for securing a network whose capabilities are accessible by external applications via an API. A request for one or more capabilities of a network is received from an application by an API Gateway of a platform that interfaces the network. Information associated with the request is input to a machine learning model to cause the machine learning model to predict whether the request is at least potentially malicious. The request is prevented from being sent to the network when the machine learning model predicts that the request is at least potentially malicious. The request is sent to the network when the machine learning model predicts that the request is not at least potentially malicious.