Bandit’s HTTP/2 header parsing accepts arbitrary octets in header field values without validation. This allows an unauthenticated remote attacker to inject raw CR, LF, or NUL characters into request headers sent over HTTP/2. The injected characters are passed directly to the application’s request header map, meaning any downstream component that consumes or logs these headers can be subverted, potentially enabling log injection, request splitting, or other header‑based attacks. Bandit itself does not provide a direct sink for the injected bytes, but the risk resides in how the application or downstream services handle the untrusted header values.

The vulnerability is present in the open‑source web framework Bandit, distributed by mtrudel, for all releases from 1.4.0 up to (but not including) 1.12.5. No other vendors or platforms are affected according to the available CNA information.

The CVSS score of 6.9 indicates moderate severity, and the advisory notes that EPS likely impacts are not publicly listed. Bandit is not included in CISA’s KEV catalog. An attacker can exploit the flaw by sending a crafted HTTP/2 HEADERS block over an open connection, after which the downstream application receives the malicious header values unchanged. Successful exploitation requires that the downstream application use the header values in a vulnerable context (e.g., unsanitized logging or request construction). The lack of a direct response‑splitting or log‑splitting sink inside Bandit itself reduces immediate risk, though the potential for downstream impact remains.