The maintainers of the open-source spam filtering system Rspamd have released version 4.3.0, addressing a series of critical security vulnerabilities and functional defects. This update resolves multiple flaws that previously allowed unauthenticated attackers to disrupt the software’s operation or bypass authentication checks. Administrators are advised to review their local configurations, as the update also introduces intentional changes to scoring mechanisms and data handling that may alter existing spam filtering verdicts.
Among the most severe security issues addressed is a crash vulnerability affecting the filter’s workers. The software was previously susceptible to service disruption via malformed HTTP requests, specifically those containing header lines missing a colon. Furthermore, the presence of an empty Deliver-To header could be exploited to repeatedly crash worker processes, effectively denying service. These bugs have been patched in the latest release to ensure greater stability against malicious input.
A separate, high-severity memory management flaw was identified within the milter proxy, the component responsible for interfacing Rspamd with mail servers. In versions 4.0.0 through 4.2.3, the proxy would incorrectly release a session twice if an upstream connection failed. This use-after-free vulnerability posed a significant risk to system integrity. The 4.3.0 release corrects this behavior, ensuring that each failed message now triggers exactly one temporary failure response, and confirms that version 3.14 remains unaffected by this specific defect.
The update also tightens the handling of DMARC and ARC authentication protocols. Previously, Rspamd incorrectly allowed ARC sets appended after a legitimate entry to weaken a DMARC rejection. The software now enforces stricter trust requirements, mandating that a forwarder must achieve a DMARC pass before the system will permit any modification to the DMARC result. Additionally, the software now properly escapes sender text before it is included in the Authentication-Results header. This prevents attackers from injecting forged authentication strings, such as dmarc=pass, into the header to deceive downstream filters.
Beyond security patches, the 4.3.0 release implements several functional changes to how Rspamd evaluates email authentication. SPF records are now processed strictly in the order of their defined terms, with redirects occurring only after all other mechanisms have been evaluated. DKIM checks have also been hardened: the system now rejects RSA keys smaller than 1024 bits, treats duplicate tags as invalid, and explicitly rejects messages with an l=0 tag if the message body is present.
Users should be aware that these updates will likely cause shifts in spam filtering scores. The system now separates URL mismatch scoring from phishing scoring, and plugins such as ASN, RBL, and DMARC have been updated to emit individual symbol options rather than comma-joined values. These changes necessitate a thorough review of any custom local scoring scripts or configurations before deployment.
Finally, the update introduces a change to how DKIM signing domains are processed. The d= values are now automatically lowercased by the system. Consequently, any existing maps or configuration files that rely on mixed-case signing domains must be updated to match this new standard. Failure to adjust these local maps could lead to unexpected behavior in signature verification.




