Cloud Crash Testing Platform Using ML-Generated App Inputs

The patent replaces the mechanical system of manual human testing with an automated machine learning-based system. The ML algorithm automatically generates inputs, executes them against the application, and detects crashes without human intervention, thereby eliminating the time-consuming manual process while maintaining or improving detection accuracy through systematic exploration of input spaces.

The system enables self-service automated testing where the machine learning algorithm independently performs the entire testing workflow - generating test inputs, executing them, monitoring application state, detecting crashes, and generating reports without requiring human testers. This autonomous operation significantly reduces the man-hours required while maintaining comprehensive testing coverage.

The patent introduces a cloud-based platform as an intermediary layer that manages the complexity of machine learning model training, execution, and coordination. This intermediary handles the sophisticated ML operations remotely, allowing the core testing system to remain relatively simple while leveraging the power of complex ML algorithms through service-based interfaces rather than direct integration.

The testing platform is designed as a universal system that can test multiple applications across different platforms and crash types using the same underlying machine learning infrastructure. By creating a multi-functional platform that handles various testing scenarios through a unified architecture, the system manages complexity through reuse and standardization rather than requiring separate complex systems for each testing need.

The system applies partial action by focusing computational resources on generating and testing only those inputs that are most likely to trigger crashes, rather than exhaustively testing all possible inputs. The machine learning algorithm learns to identify high-value test cases that provide maximum crash detection coverage with minimal computational expenditure, achieving reliable detection without excessive resource consumption.

The system performs preliminary action by pre-training machine learning models on historical crash data and application behavior patterns before actual testing begins. This preliminary training enables the system to make intelligent predictions about which inputs are likely to cause crashes, allowing it to achieve comprehensive detection coverage more efficiently during the actual testing phase without needing to exhaustively explore all input possibilities.