Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-105147 Overview
CVE-2026-105147 is a hard-coded credentials vulnerability [CWE-259] in SciPhi-AI R2R versions up to 3.6.6. The flaw resides in the JWT Secret Handler component, where the DEFAULT_BCRYPT_SECRET_KEY and DEFAULT_NACL_SECRET_KEY values are embedded in the code. Attackers can leverage these predictable secrets remotely without authentication. A public proof-of-concept has been disclosed, increasing exposure for unpatched deployments. The vendor was contacted prior to publication but did not respond, meaning no coordinated fix is confirmed at the time of disclosure.
Critical Impact
Hard-coded JWT secrets allow remote attackers to forge valid authentication tokens, bypass access controls, and impersonate any user of the R2R application.
Affected Products
- SciPhi-AI R2R versions up to and including 3.6.6
- Deployments using default DEFAULT_BCRYPT_SECRET_KEY configuration
- Deployments using default DEFAULT_NACL_SECRET_KEY configuration
Discovery Timeline
- 2026-10-04 - CVE-2026-105147 published to NVD
- 2026-10-05 - Last updated in NVD database
Technical Details for CVE-2026-105147
Vulnerability Analysis
SciPhi-AI R2R ships with static default values for DEFAULT_BCRYPT_SECRET_KEY and DEFAULT_NACL_SECRET_KEY. These keys are used by the JWT Secret Handler to sign and verify authentication tokens and to protect secrets at rest. When operators deploy R2R without overriding these constants, the application signs tokens with a value that is publicly visible in the source repository.
An attacker who knows the hard-coded secret can craft a JWT with arbitrary claims, including administrative identities. The R2R backend will validate the signature against the same embedded key and accept the forged token. This converts an unauthenticated remote request into a fully authenticated session.
The issue is network-exploitable, requires no privileges, and no user interaction. A public gist contains working proof-of-concept code demonstrating token forgery.
Root Cause
The root cause is a design-level use of hard-coded cryptographic secrets as fallback values rather than requiring operators to generate unique keys at install time. The constants persist in the shipped binary and configuration defaults, so any deployment that does not explicitly override them inherits the same key material as every other installation.
Attack Vector
Exploitation proceeds in three steps. The attacker retrieves DEFAULT_BCRYPT_SECRET_KEY and DEFAULT_NACL_SECRET_KEY from the public source. The attacker then uses the key to sign a JWT carrying a chosen user_id or role claim. Finally, the attacker submits the forged token to any R2R API endpoint that enforces JWT authentication, gaining access as the impersonated principal.
For technical details, see the GitHub Gist PoC Code and VulDB CVE-2026-105147.
Detect Identity Attacks the Moment Authentication Fails
Singularity™ Identity surfaces credential misuse, lateral movement, and privilege escalation in real time, so you can contain the breach before it spreads.
Stop Identity AttacksDetection Methods for CVE-2026-105147
Indicators of Compromise
- JWT tokens in access logs signed with the known default DEFAULT_BCRYPT_SECRET_KEY or DEFAULT_NACL_SECRET_KEY values
- Authentication events for high-privilege accounts originating from unexpected IP addresses or user agents
- API requests to R2R endpoints that bypass the normal login or OAuth flow but present a valid bearer token
- Session activity on R2R instances still running version 3.6.6 or earlier without secret rotation
Detection Strategies
- Inspect R2R configuration and environment variables for the presence of the shipped default secret values
- Correlate JWT iat and exp claims against expected session issuance from the authentication service
- Alert on token claims containing elevated roles that were not issued through the standard login workflow
- Monitor for anomalous document ingestion, retrieval, or administrative API calls on R2R deployments
Monitoring Recommendations
- Centralize R2R application logs and authentication events in a SIEM for retrospective analysis
- Track configuration drift so that any revert to default secret values triggers an alert
- Baseline normal API usage patterns per user and flag deviations in request rate or endpoint mix
How to Mitigate CVE-2026-105147
Immediate Actions Required
- Replace DEFAULT_BCRYPT_SECRET_KEY and DEFAULT_NACL_SECRET_KEY with high-entropy, randomly generated values unique to each deployment
- Invalidate all existing JWTs issued under the default keys and require users to re-authenticate
- Restrict network exposure of R2R management APIs to trusted networks or VPN segments
- Review access logs for signs of forged tokens or unauthorized administrative activity
Patch Information
At the time of CVE publication, the vendor had not responded to disclosure and no official patch is referenced. Operators should monitor the SciPhi-AI R2R repository for updates and apply fixed releases as soon as they become available. Until then, manual secret rotation is the primary remediation.
Workarounds
- Set DEFAULT_BCRYPT_SECRET_KEY and DEFAULT_NACL_SECRET_KEY to unique secrets via environment variables or a secrets manager before starting the service
- Place R2R behind an authenticating reverse proxy that enforces mutual TLS or an additional bearer token layer
- Rotate secrets on a scheduled cadence and after any suspected compromise
- Disable or firewall off R2R instances that cannot be reconfigured until a patched version is available
# Configuration example: generate and export unique secrets before launch
export DEFAULT_BCRYPT_SECRET_KEY="$(openssl rand -base64 48)"
export DEFAULT_NACL_SECRET_KEY="$(openssl rand -base64 48)"
# Verify no default values remain in the running configuration
env | grep -E 'BCRYPT_SECRET_KEY|NACL_SECRET_KEY'
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

