Method, system and product for using a predictive model to predict if inputs reach a vulnerability of a program
US10387655B2 · kind B2 · utility
Assignee
Inventors
Key dates
| Filing date | Feb 15, 2017 |
| Grant date | Aug 20, 2019 |
| Priority date | — |
| Expiry date | Nov 10, 2037 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06N7/023
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
A method, apparatus and product for using a predictive model to predict if inputs reach a vulnerability of a program. Given a sample input, which when provided to a program being executed, is configured to cause execution of the program to reach a vulnerability, a set of variant inputs and labels thereof is generated based on the sample input and execution of the program with each variant input. A predictive model is trained based on the set of variant inputs and labels thereof, and provided to an input analysis platform configured to analyze an input; whereby the input analysis platform is enabled to predict whether an input would cause the program to reach the vulnerability prior to executing the program with the input and perform a responsive action accordingly.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.