Browser sample set acquisition method based on deep learning and genetic algorithm

Abstract

The invention discloses a browser fuzzy test sample set acquisition method. The method comprises the following steps of (1) preprocessing a document, and counting html file elements to obtain an inputvector and element statistical data; and (2) performing deep learning by using an LSTM neural network, and decoding the generated sample to obtain a generated sample. (3) performing tree coding on the generated sample to obtain a parent population, and calculating a fitness function; and (4) optimizing the parent population by using the optimized genetic operator until a termination result is met. The method can be used for browser vulnerability mining, the mining direction is more targeted, and mining efficiency is higher.

Description

technical field [0001] The invention relates to a browser sample set acquisition method based on deep learning and genetic algorithm, belonging to the technical field of vulnerability mining. Background technique [0002] As a necessary tool for surfing the Internet, the browser occupies a very important position, and its frequency of security threats and attacks is higher than that of other software. In common attack scenarios, attackers attack browsers or other files containing links to make browsers parse data incorrectly, obtain cached data by attacking browser memory, or use browsers as springboards to analyze communication protocols, attack servers, and databases. The memory leaked by the browser may include sensitive information such as personal account numbers and passwords, and it is possible to obtain a large amount of user data by attacking the browser server and database. The vulnerability mining of the browser can expose the security risks of the browser in adv...

AI Technical Summary

Problems solved by technology

At present, this method also faces three problems. One is to perform specific optimization for known samples, and cannot universally mine specific types of vulnerabilities.

The second is that the protection measures in the target program cannot be bypassed, and computing resources are wasted when loops or repeated calls are encountered

However, at present, there is no general process and framework for the generation and optimization of the vulnerability mining sample set of the browser itself.

Images