Skip to content

aiosmtplib: STARTTLS response injection

Moderate severity GitHub Reviewed Published Jun 20, 2026 in cole/aiosmtplib • Updated Aug 27, 2026

Package

pip aiosmtplib (pip)

Affected versions

<= 5.1.1

Patched versions

5.1.2

Description

Impact

When a connection is upgraded with STARTTLS, aiosmtplib reads the server's 220 go-ahead reply and immediately performs the TLS handshake without discarding any data still sitting in the receive buffer. Bytes the protocol read off the plaintext socket before the handshake survive across the plaintext→TLS boundary (the asyncio transport is swapped in place, so the protocol object and its buffer are reused), and are then parsed as though they had arrived inside the TLS session.

Who is affected: Any caller that uses STARTTLS by passing start_tls=True or start_tls=None when the server advertises STARTTLS, and whose traffic can be intercepted by an active network attacker on the plaintext leg of the connection.

A man in the middle can send, in a single segment immediately after the client's STARTTLS command, the 220 reply followed by attacker-chosen response lines (e.g. 220 Go ahead\r\n250-mx.evil\r\n250 AUTH LOGIN\r\n). aiosmtplib
consumes only the 220, leaves the injected lines buffered, completes the handshake, and then parses the attacker's pre-staged plaintext as the first post-TLS server response. This also desynchronizes every
subsequent command/response pair inside the "encrypted" session.

Not affected: Connections using implicit/direct TLS (use_tls=True) have no plaintext phase and are not vulnerable. The attack requires an active man in the middle via network compromise; a passive eavesdropper cannot exploit it.

Patches

A fix is available in aiosmtplib 5.1.2. All earlier versions that support STARTTLS are affected; upgrade to 5.1.2 or
later.

The fix treats any data buffered after the 220 STARTTLS reply and before the handshake as a protocol violation, per RFC 3207 §4.2 ("the client MUST discard any knowledge obtained from the server … which was not
obtained from the TLS negotiation itself").

Workarounds

If you cannot upgrade immediately:

  • Use implicit TLS instead of STARTTLS. Connect with use_tls=True. This removes the plaintext phase entirely.
  • If STARTTLS is unavoidable, restrict connections to servers reached over a trusted network path.

References

@cole cole published to cole/aiosmtplib Jun 20, 2026
Published by the National Vulnerability Database Aug 20, 2026
Published to the GitHub Advisory Database Aug 27, 2026
Reviewed Aug 27, 2026
Last updated Aug 27, 2026

Severity

Moderate

CVSS overall score

This score calculates overall vulnerability severity from 0 to 10 and is based on the Common Vulnerability Scoring System (CVSS).
/ 10

CVSS v3 base metrics

Attack vector
Network
Attack complexity
High
Privileges required
None
User interaction
None
Scope
Unchanged
Confidentiality
None
Integrity
High
Availability
None

CVSS v3 base metrics

Attack vector: More severe the more the remote (logically and physically) an attacker can be in order to exploit the vulnerability.
Attack complexity: More severe for the least complex attacks.
Privileges required: More severe if no privileges are required.
User interaction: More severe when no user interaction is required.
Scope: More severe when a scope change occurs, e.g. one vulnerable component impacts resources in components beyond its security scope.
Confidentiality: More severe when loss of data confidentiality is highest, measuring the level of data access available to an unauthorized user.
Integrity: More severe when loss of data integrity is the highest, measuring the consequence of data modification possible by an unauthorized user.
Availability: More severe when the loss of impacted component availability is highest.
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:H/A:N

EPSS score

Exploit Prediction Scoring System (EPSS)

This score estimates the probability of this vulnerability being exploited within the next 30 days. Data provided by FIRST.
(34th percentile)

Weaknesses

Improper Neutralization of Special Elements in Output Used by a Downstream Component ('Injection')

The product constructs all or part of a command, data structure, or record using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify how it is parsed or interpreted when it is sent to a downstream component. Learn more on MITRE.

CVE ID

CVE-2026-55558

GHSA ID

GHSA-vxj7-4xrp-5vr4

Source code

Loading Checking history
See something to contribute? Suggest improvements for this vulnerability.