Bouncy Castle C# CVE-2026-103600

Exploitation requires that the target application use bc-csharp before 2.7.0 to parse attacker-supplied ASN.1 data through any path that reaches Asn1InputStream/Asn1StreamParser - including X.509 certificate/CRL validation, CMS/PKCS#7, PKCS#8/PKCS#12, OCSP, or TLS Certificate-message handling - with no authentication or user interaction needed (PR:N/UI:N). The crafted input needs only ~2,000 levels of nested constructed elements (~8 KB DER, or indefinite-length BER) to reach the default depth. The outright process-killing crash requires the parsing thread to run on the ~1.5 MB default stack (the .NET main-thread default on Windows); on threads configured with larger stacks the effect is instead a quadratic CPU slowdown (about 9 seconds for 64 KB) rather than an immediate StackOverflowException. The practical limiting factor is therefore thread stack size and whether the deployment exposes untrusted ASN.1 parsing to remote input.

This is a credible, actionable availability risk rather than a high-CVSS-but-low-real-risk issue. The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:N) confirms network-reachable, low-complexity, unauthenticated exploitation with no user interaction, and the impact is confined to availability (VA:H; VC:N/VI:N) - consistent with a DoS and CWE-674 (Uncontrolled Recursion). The severity is amplified by the failure mode: a StackOverflowException in .NET cannot be caught and terminates the entire process, so a single ~8 KB crafted input can take down a service outright on default Windows main-thread stacks; on larger stacks the impact degrades to quadratic CPU consumption (~9s for 64 KB) rather than a crash. No EPSS score, KEV listing, or POC was provided in the input, so predicted and confirmed exploitation signals are absent - exploitation status is 'no public exploit identified at time of analysis.' However, absence of those signals should not be read as low risk: the vendor (Bouncy Castle) has published an advisory and shipped a fix, the bug is trivially reproducible, and the exposed attack surface (any untrusted ASN.1: X.509 certs/CRLs, CMS/PKCS#7, PKCS#8/PKCS#12, OCSP, TLS Certificate messages) is reached by routine TLS and PKI operations. There are no conflicting signals: the network vector in the CVSS vector matches the description's remote/untrusted-input premise.

An attacker sends a crafted ASN.1 structure - roughly 2,000 nested SEQUENCE-in-SEQUENCE elements in about 8 KB of DER/BER - to any endpoint of an affected .NET application that parses untrusted ASN.1, such as a TLS server/client processing a Certificate message, or a service validating a submitted X.509 certificate or CMS/PKCS#7 blob. The recursive parser descends one call per nesting level with no depth bound, exhausting the ~1.5 MB default thread stack and raising an uncatchable StackOverflowException that crashes the whole process. No public exploit was identified at time of analysis, but the technique is simple and fully described in the advisory.