pyasn1 EUVD-2026-113352

The affected application must pass untrusted, attacker-sized ASN.1 to a pyasn1 decoder path that reaches constructed BIT STRING assembly: the BER and CER decoders reach it with or without an asn1Spec; the DER decoder is exposed ONLY on its schema-driven path when given asn1Spec=univ.BitString() and fed a definite-length constructed BIT STRING (constructed BIT STRING is not valid DER - this is decoder permissiveness). Conforming CER with indefinite-length outer framing and 1000-octet primitive fragments also triggers the indefinite-length loop. Limiting factors: schema-less DER decoding REJECTS the constructed value and is not vulnerable; the attacker needs enough fragments that the application's input limits still admit (a pre-decode input-size cap bounds, though does not eliminate, exposure); and actual network reachability and authentication requirement are determined by the calling application, not by pyasn1 itself. Impact is CPU exhaustion only - no data exposure or modification.

This is a genuine but conditional availability risk, not a top-tier emergency. The library-level CVSS of 7.5 (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H) reflects a worst-case assumption stated explicitly in the GHSA advisory: a 'reasonable application that automatically decodes attacker-supplied ASN.1 from unauthenticated network input.' The CWE-407 (quadratic/algorithmic complexity) classification and the confirmed O(n²) behavior - benchmarked at 7.4 s CPU for a 0.8 MB BER input on v0.6.4, with an independent report of 21.1 s for 1.28 MB - substantiate the A:H impact. However, the real-world risk is entirely gated by the calling application: the advisory states 'Whether the input is network-accessible or unauthenticated depends on the calling application' and 'makes no claim that every application using pyasn1 ... accepts such input.' The vuln.today Decision label is 'this_week' (not act_now/emergency), consistent with a serious issue that is not under active mass exploitation. There is NO CISA KEV listing and NO EPSS score provided, so exploitation is not confirmed in the wild - appropriate for a DoS primitive that has no confidentiality or integrity impact and must be chained through a specific application entry point. Confidentiality and integrity are N, so this cannot be escalated beyond denial of service. Net: prioritize patching where pyasn1 decodes untrusted ASN.1 from the network, but the raw 7.5 overstates risk for deployments that only decode trusted/bounded input.

An attacker who can submit ASN.1 data to an application that automatically decodes it with pyasn1's BER or CER decoder sends a small (sub-megabyte) payload containing tens of thousands of constructed BIT STRING fragments. During decode(), pyasn1 repeatedly shifts and copies an ever-growing integer per fragment, consuming quadratically increasing CPU time and tying up the decoding thread or worker pool; repeated requests sustain a denial of service. A working reproducer is published in the GHSA advisory demonstrating the quadratic cost on v0.6.4.