SymGraph

On this page

Workflow: Building a Semantic Graph

This guide walks you through building and using a semantic graph of your binary, enabling rich contextual understanding and security analysis.

Overview

The Semantic Graph creates a knowledge base of your binary that captures:

This graph enhances LLM queries and enables structured exploration of the binary.

When to Build a Semantic Graph

Build a semantic graph when you want to:

Step-by-Step Workflow

Step 1: Open the Semantic Graph Tab

Navigate to the Semantic Graph tab in BinAssist. You’ll see the List View sub-tab with an empty or previously populated function list.

Step 2: ReIndex the Binary

Click ReIndex Binary in the Manual Analysis panel to build the foundation.

This process: 1. Extracts all functions from the binary 2. Records function addresses, names, and sizes 3. Builds the call graph (CALLS edges) 4. Identifies imports and exports

The time required depends on binary size: - Small binaries (< 100 functions): A few seconds - Medium binaries (100-1000 functions): Under a minute - Large binaries (1000+ functions): Several minutes

Step 3: Run Semantic Analysis

Click Semantic Analysis to generate LLM-powered summaries.

For each function, the LLM generates: - Purpose summary - Activity profile (what the function does) - Security flags (network, file I/O, crypto, etc.) - Risk assessment

This step is LLM-intensive and may take time for large binaries. Progress is shown in the status area.

Step 4: Run Security Analysis

Click Security Analysis to perform taint analysis.

This process: 1. Identifies sources (functions that receive external input) - Network receive functions (recv, read from socket) - File read functions - User input functions 2. Identifies sinks (potentially dangerous operations) - strcpy, sprintf (buffer overflows) - system, exec (command injection) - SQL functions (injection) 3. Traces paths from sources to sinks 4. Creates TAINT_FLOWS_TO and VULNERABLE_VIA edges

Discovered vulnerability paths are stored in the graph for review.

Step 5: Run Community Detection

Click Community Detection to group related functions.

Using the Label Propagation algorithm, this: 1. Analyzes call relationships 2. Groups functions that frequently interact 3. Labels communities by common purpose

Communities help you understand the binary’s modular structure: - “Network I/O” - functions handling network communication - “Crypto” - cryptographic operations - “File Operations” - file system interactions - “String Processing” - string manipulation functions

Exploring the Graph

List View

Browse all indexed functions: - Click a row to see callers and callees - Sort by clicking column headers - Use the table to navigate the binary

Visual Graph

Explore relationships visually: - Nodes represent functions - Edges show call relationships - Colors indicate communities or security flags - Click nodes for details

Visual Graph

Find functions by keyword: 1. Switch to the Search sub-tab 2. Enter keywords (function names, summary terms) 3. Review results with relevance scores

Using the Graph with LLM Queries

The semantic graph enhances LLM queries when MCP is enabled:

  1. Build the semantic graph (Steps 1-5 above)
  2. Go to the Query tab
  3. Enable MCP
  4. Ask questions that benefit from graph context

The LLM can now: - Query for functions by purpose - Find related functions - Trace call chains - Identify security-relevant code

Example queries: - “What functions handle network input?” - “Show me the call chain from main to any crypto functions” - “Which functions have high security risk?”

Incremental Updates

You don’t need to rebuild everything when the binary changes:

Tips for Effective Graph Building

Start Small

For large binaries: 1. ReIndex to get the structure 2. Run Semantic Analysis on key functions first 3. Expand analysis as needed

Focus on Interesting Areas

Not every function needs deep analysis: - Entry points and exports - Functions with security flags - Functions in critical call paths

Iterate

The graph is a living resource: 1. Build initial graph 2. Use it in queries 3. Refine based on what you learn 4. Re-run analyses as understanding deepens

Combine with Manual Analysis

Use the graph alongside manual reverse engineering: - Graph shows the big picture - Manual analysis provides depth - Queries connect the two