Security Protection Against Side-Channel Attacks In Browsers

With the rapid development of Internet technology,more and more people rely on Web browsers to meet their life and learning needs.With the gradual enrichment of browser functions,the browser’s internal information implies a large amount of privacy information,and user’s usage habits and other behavioral characteristics,which poses challenges to the security of user information.Side-channel attacks usually use subtle differences generated by different running states of the program to infer user secret information.The JavaScript code in the website can be executed remotely in the restricted environment of the browser sandbox,so side-channel attacks based on JavaScript are gradually gaining attention.Attackers use Java Scipt to collect specific information about the browser itself,and use them to uniquely identify the browser,further identify the user,and even identify other websites that are being visited.Moreover,the attacker can bypass the JavaScript sandbox to detect the layout of the memory,and even read any information in the memory out of bounds and trigger a memory bit flip.How to defend against side-channel attacks in the browser has become an important issue to be solved in the field of information security.Based on the internal attributes of the browser,this paper proposes a side-channel defense scheme in browser based on sensitive features.Sensitive features are defined as browser attributes that differ in different environments and can therefore be exploited to reveal user and environmental information.Firstly,we propose an algorithm to acquire and filter sensitive features.According to the structure and characteristics of JavaScript in the browser,all accessible attributes are obtained.After automatic filtering and manual semantic analysis,a sensitive feature set is obtained.Then we quantize how many times the sensitive features are called in normal websites,and analyze the result and set three impact levels accordingly.Finally,we implement the threat mitigation solution as a browser extension,and the idea of a wrapper is used inside the extension.According to the impact level of each sensitive feature,there are three different degrees of packaging and replacement schemes called “block”,“modify”,and “add”.The main research contents and innovations of this article are as follows:1)An algorithm to acquire and filter sensitive features.Firstly,use JavaScript to obtain all the accessible attributes,including all object attributes that can be directly accessed,traversed from the window object,and manually added artificial attributes about Web GL.Then,by filtering out repeated attributes and non-static attributes in a single environment,the effective attribute set in the current environment is obtained.Finally,the indistinguishable attributes are filtered out by comparing the effective attribute sets of multiple environments offline,and then manually analyzed the rest features to further filter out forty sensitive features.2)Analyze and quantify the use of sensitive feature sets in normal websites,and obtain the impact level of each sensitive feature on the website,and adopt different degrees of replacement schemes for feature values according to different impact levels.Firstly,obtain the JavaScript code contained in the Top 100 websites in Alexa,and then obtain the number of calls of each feature in the sensitive feature set through static analysis,so as to quantify the impact of modifying the sensitive feature on the normal website into low-level,intermediate-level and high-level.Take “block”,“modify”,and“add” operations for the three levels of sensitive features.3)A Chrome extension is implemented using the idea of wrapper.Package and replace sensitive features with different impact levels within the extension.The lowlevel corresponds to the "block" operation,which sets the feature to unaccessible and returns undefined directly.The intermediate-level corresponds to the "modify" operation,which will replace the original value of the attribute with a random value inside the extension and return.High-level corresponds to the "add" operation.The feature value will not be modified,but some random attributes will be added to the browser,thereby interfering with the attacker’s identification.4)Results analysis of the algorithm and performance evaluation of the extension.Analysis of the algorithm results shows that the screened sensitive features can indeed be used to identify user and environmental information.An analysis of the effectiveness of the extension shows that the extension can effectively prevent JavaScript sidechannel attacks.The performance evaluation of the extension shows that the overhead of enabling the extension is within the normal range and will not have any negative impact on normal operation.