sos
stack overflow security scanning
approach
Research Approach We will be focusing on discovering patterns with vulnerabilities found in StackOverflow. As not only an open-source forum and aid for almost every developer worldwide, we wish to investigate the integrity and security strength of community feedback. We suspect there may be higher probabilities of vulnerable code within certain topics, such as with coding languages like PHP or C++. Though there are many classifications of a security vulnerability, we will focus on using static analysis in code snippets. Through a brief inspection of our available datasets (Google BigQuery), our first stage will be to determine the assigned tags based on the question asked. As the dataset is missing labels/tags, we will be using StackOverflow’s application and API to retrieve the tags. This will help us categorize the answers to discover patterns if certain topics are more prone to vulnerabilities. We will be setting up a database and running custom scripts to make requests to StackOverflow’s servers. We currently estimate that given 19 million unique questions and a rate of 30 requests per second, the labeling process would take about 7.3 days. Most likely we will also implement some distributed systems (or cloud service) to help speed up that process. The second stage will be to scan the answers using semgrep, a static application security testing (SAST) tool. Even though a StackOverflow answer may consist of a description with a small code snippet, we hypothesize that developers may not understand the full concept of an answer and will likely copy-paste the snippet into their code. semgrep also comes with its own registry, providing parameters to detect security vulnerabilities. Given that semgrep is written in OCaml, we estimate the scanning process will not take as much time compared to stage one. The main question we will determine is how to set up the scanning as well as categorize possible severities of detections. False positives may also be an issue as well.