hdds_viewer Network Analyzer

hdds_viewer is a military-grade observability tool for DDS/RTPS networks. It provides real-time capture, ML-powered anomaly detection, and visual debugging for robotics, defense, automotive, and medical applications. HDDS Viewer Main Interface

Key Features

Core Functionality

Offline replay from .hddscap capture files (625 MB/s I/O throughput)
PCAP import - Analyze existing network captures
Message inspection with CDR2 decoding (14 primitive types)
Topology visualization (force-directed graph + list views)
Timeline analysis - Latency tracking with p50/p95/p99 percentiles

Intelligence

QoS Advisor - 15 automated DDS misconfiguration detection rules
AI Assistant - 15 pattern detectors for anomaly detection with suggested fixes
ML Detection - ONNX-powered inference with 99.58% accuracy

User Experience

8 themes - Dark Pro, Cyberpunk, High Contrast (WCAG AAA), Catppuccin variants, Gruvbox
5 languages - English, French, German, Italian, Japanese (auto-detect)
Plugin API - Extend with custom analyzers (Enterprise)
Multi-format export - JSON, CSV, YAML, Markdown

Quick Start

# Install via cargo cargo install hdds-viewer Open a capture file (GUI mode) hdds-viewer capture.hddscap Headless analysis with JSON output hdds-viewer --analyze capture.hddscap --format json Run with ML-powered detection

hdds-viewer --analyze capture.hddscap --ml-detect

Architecture

hdds_viewer is built as a Rust workspace with multiple crates:

Crate

Purpose

`viewer-core`	Capture I/O, CDR decoding, topology, QoS analysis
`viewer-ui`	egui/eframe GUI application
`viewer-ai`	Pattern detectors (15 rules) and LLM integration
`viewer-ml`	ONNX Runtime inference and feature extraction
`viewer-live`	Live DDS capture via HDDS Telemetry TCP
`viewer-plugin`	Plugin API for custom analyzers

CLI Options

hdds-viewer [OPTIONS] [FILE]

Arguments:
  [FILE]  Path to .hddscap or .pcap file to analyze

Options:
  -a, --analyze          Run headless analysis (no GUI)
  -f, --format <FMT>     Output format: json, yaml, csv, text (default: text)
  --ml-detect            Enable ML-powered anomaly detection
  --ml-threshold <N>     Anomaly confidence threshold (default: 0.85)
  --ai                   Enable LLM-powered analysis (requires backend)
  --dump-features        Export ML features to CSV (for model training)
  --no-ml                Disable ML detection
  --no-network           Air-gapped mode (no external connections)
  -h, --help             Print help
  -V, --version          Print version

Exit Codes

Code

Meaning

0	No issues detected
1	Low severity anomalies
2	Medium severity anomalies
3	High severity anomalies
4	Critical issues or file errors

Capture Formats

Format

Extension

Description

HDDSCAP	`.hddscap`	Native format, optimized for ML analysis
PCAP	`.pcap`, `.pcapng`	Wireshark-compatible network capture

Detection Systems

hdds_viewer includes three complementary detection systems:

1. QoS Advisor (15 rules)

Detects DDS/RTPS misconfigurations:

Core QoS	Reliability, Durability, History, Deadline, Lifespan
Extended QoS	Ownership, Partition, Latency Budget, Transport Priority
Resources	Resource Limits, Liveliness, Content Filter

2. AI Assistant (15 pattern detectors)

Automated anomaly detection with suggested fixes:

Severity

Detectors

Critical	Cascade failure, Resource exhaustion, Liveliness timeout
High	Deadline violation, Ownership conflict, Type mismatch
Medium	GC spikes, Partition mismatch, Presentation order
Low	Periodic patterns, Content filter efficiency

See AI Assistant documentation for details.

3. ML Detection (ONNX)

Machine learning inference with ensemble model:

Metric

Value

Accuracy	99.58%
Inference latency	< 2ms per 100 frames
Features	50+ statistical features

See ML Detection documentation for details.

GUI Panels

Panel

Purpose

Frame Table	Browse messages with filtering, sorting, export
Topology	Interactive force-directed graph + list view
Timeline	Latency analysis with p50/p95/p99 percentiles
QoS Advisor	Configuration issue detection with fixes
AI Assistant	Pattern-based anomaly detection
Metrics	Throughput charts + latency histograms
Inspector	Payload hex/JSON/raw view
Recorder	Record new `.hddscap` captures

Topology View

Timeline & Metrics

Message Inspector

Keyboard Shortcuts

Shortcut

Action

Space	Play/Pause playback
Ctrl+O	Open capture file
Ctrl+R	Restart playback
Ctrl+Q	Open QoS Advisor
Ctrl+A	Open AI Assistant
Ctrl+,	Open Settings
Esc	Close dialog/window

Performance

Metric

Value

Write throughput	625 MB/s
Read throughput	800+ MB/s
RingBuffer ingestion	41M msg/s
UI framerate	60 FPS at all message rates
Binary size	12 MB

Adaptive Sampling

Message Rate

Sampling

< 30k msg/s	Full (all frames)
30k-60k msg/s	1/2 (every 2nd frame)
60k-100k msg/s	1/4 (every 4th frame)
> 100k msg/s	1/10 (every 10th frame)

Configuration

Config File

Edit ~/.config/hdds-viewer/config.toml:

[default]
domain = 0
format = "pretty"

[appearance]
theme = "dark-pro"
language = "en"
ui_scale = 1.0

[ml]
enabled = true
anomaly_threshold = 0.85
model_path = "~/.config/hdds-viewer/models/"

[ai]
backend = "ollama"
model = "llama3.2"
base_url = "http://localhost:11434"

Environment Variables

HDDS_VIEWER_DOMAIN=0       # Default DDS domain
HDDS_VIEWER_THEME=cyberpunk # UI theme
HDDS_VIEWER_LANG=fr        # Language (en, fr, de, it, ja)
HDDS_VIEWER_LOG_LEVEL=debug # Log verbosity

System Requirements

Requirement

Minimum

OS	Debian 11+, Ubuntu 20.04+ (glibc 2.31+)
Graphics	OpenGL 3.3+ or Vulkan
Memory	512 MB (+ capture file size)
Disk	50 MB (+ models ~5 MB)

Quality Standards

hdds_viewer enforces military-grade quality (ANSSI/IGI-1300, Common Criteria EAL4+):

Zero CRITICAL/HIGH audit violations
All unsafe blocks require // SAFETY: comments
No production panics outside tests
164 tests (126 unit + 23 integration + 12 doc + 3 tool)

License

Community Edition: Free for personal/non-commercial use
Pro/Enterprise: Commercial license with additional features
Offline validation, no telemetry