CVE-2025-59354: Dragonfly Hash Collision Vulnerability

CVE-2025-59354 Overview

Dragonfly is an open source peer-to-peer (P2P) based file distribution and image acceleration system maintained under the Linux Foundation. Prior to version 2.1.0, Dragonfly relies on weak hash functions, including MD5, to verify integrity of downloaded files. Attackers can substitute legitimate files with malicious payloads that produce colliding hash values, bypassing integrity checks in the distribution pipeline. The weakness is cataloged under [CWE-328] (Use of Weak Hash) and affects file integrity guarantees across the P2P mesh. The vendor released a fix in Dragonfly 2.1.0.

Critical Impact
Attackers can replace distributed files with malicious content that produces identical MD5 hashes, undermining integrity verification across the Dragonfly P2P network.

Affected Products

Linux Foundation Dragonfly (linuxfoundation:dragonfly) versions prior to 2.1.0
Dragonfly2 Go implementation (cpe:2.3:a:linuxfoundation:dragonfly:*:*:*:*:*:go:*:*)
Container image acceleration deployments using vulnerable Dragonfly releases

Discovery Timeline

2025-09-17 - CVE-2025-59354 published to NVD
2026-06-17 - Last updated in NVD database

Technical Details for CVE-2025-59354

Vulnerability Analysis

Dragonfly distributes files across a peer-to-peer network and uses hash functions to validate the integrity of transferred content. Prior to release 2.1.0, the codebase includes MD5 among the algorithms used for file verification. MD5 has documented collision weaknesses that allow attackers to craft two distinct files producing identical digest values.

An attacker positioned to inject or serve alternate content within the distribution mesh can supply a malicious file whose MD5 digest matches the expected value for a legitimate artifact. The Dragonfly client accepts the substituted content as valid, breaking the guarantee that peers receive authentic files.

The issue targets the confidentiality and integrity of the software supply chain path that Dragonfly accelerates, particularly container image and large-file distribution scenarios used in Kubernetes and cloud-native workloads.

Root Cause

The root cause is the continued use of a cryptographically broken hash algorithm ([CWE-328]) for security-relevant integrity verification. MD5 collision attacks have been demonstrated publicly for years, so any use of MD5 as a trust anchor for file content is unsafe. Dragonfly did not enforce a collision-resistant alternative such as SHA-256 for all verification paths prior to 2.1.0.

Attack Vector

Exploitation requires network access to the Dragonfly distribution flow. An attacker prepares two files—one benign, one malicious—that share the same MD5 digest through a chosen-prefix collision. When Dragonfly peers request the expected artifact, the attacker serves the malicious variant. Integrity checks pass because the MD5 digest matches the registered value. No authentication or user interaction is required for the substitution to succeed against the verification layer.

The vulnerability manifests in the file verification routines used by Dragonfly2 peers. See the GitHub Security Advisory GHSA-hx2h-vjw2-8r54 for the vendor's technical description.

Detection Methods for CVE-2025-59354

Indicators of Compromise

Unexpected file content served by Dragonfly peers despite matching hash metadata
Container images or artifacts downloaded through Dragonfly showing behavior inconsistent with the upstream registry version
Log entries from Dragonfly dfget or dfdaemon components showing repeated retrieval from unusual peer IPs
Discrepancies between MD5 digests reported by Dragonfly and SHA-256 digests computed independently on the same artifact

Detection Strategies

Compute SHA-256 (or stronger) digests of files retrieved via Dragonfly and compare against upstream registry values published by the source of truth
Inventory Dragonfly deployments and flag any peer or scheduler running versions earlier than 2.1.0
Correlate container image pull events with subsequent runtime process anomalies to identify substituted artifacts
Monitor Dragonfly configuration for use of md5 as the digest algorithm in task or piece verification

Monitoring Recommendations

Ingest Dragonfly dfdaemon and scheduler logs into a centralized log platform for continuous review
Alert on version banners or Prometheus metrics indicating Dragonfly builds below 2.1.0
Track process execution and file writes originating from binaries or images distributed via Dragonfly to detect tampering post-delivery

How to Mitigate CVE-2025-59354

Immediate Actions Required

Upgrade all Dragonfly components (manager, scheduler, dfdaemon, seed peer) to version 2.1.0 or later
Audit recent artifact distributions for integrity by recomputing SHA-256 hashes against trusted upstream sources
Restrict network reachability of Dragonfly peers to trusted management and cluster networks

Patch Information

The vulnerability is fixed in Dragonfly 2.1.0. Refer to the vendor advisory GHSA-hx2h-vjw2-8r54 for release notes and upgrade guidance. Additional context on Dragonfly's security posture is available in the Dragonfly Comprehensive Security Report 2023.

Workarounds

Configure Dragonfly to use SHA-256 or stronger digest algorithms in task verification where the configuration exposes that option
Perform out-of-band integrity verification using SHA-256 or SHA-512 for critical artifacts before deployment
Limit peer-to-peer traffic to authenticated internal networks and disable unnecessary public-facing Dragonfly endpoints

bash

# Upgrade Dragonfly to the patched release
helm repo update
helm upgrade dragonfly dragonfly/dragonfly \
  --namespace dragonfly-system \
  --version 2.1.0

# Verify running versions
kubectl get pods -n dragonfly-system -o jsonpath='{.items[*].spec.containers[*].image}'