RPKI infrastructure attacks¶
Attack pattern¶
RPKI is the cryptographic answer to prefix hijacking: sign who may originate which prefix, and let routers reject the rest. But the defence is itself a system, a hierarchy of registries and certificate authorities, repositories fetched over rsync and the RRDP delta protocol, validators feeding routers over the RPKI-to-Router (RTR) protocol, and every one of those is a thing that can be broken, delayed, or lied to. Attack the infrastructure that proves routes honest and the routes stop being proven, which puts hijacking back on the table without touching BGP at all.
1. RPKI infrastructure attacks [OR]
1.1 Repository system attacks [OR]
1.1.1 Publication point manipulation
1.1.1.1 RPKI object deletion or modification
1.1.1.2 Stale object substitution attacks
1.1.1.3 Man-in-the-middle attacks on repository transfers
1.1.2 Repository service exploitation
1.1.2.1 RSYNC server vulnerabilities
1.1.2.2 RRDP (RPKI Repository Delta Protocol) implementation flaws
1.1.2.3 Denial-of-service against repository servers
1.1.3 Content delivery network attacks
1.1.3.1 CDN cache poisoning for RPKI objects
1.1.3.2 HTTPS interception and manipulation
1.1.3.3 DNS spoofing for repository locations
1.2 Cryptographic attack vectors [OR]
1.2.1 Certificate authority compromise
1.2.1.1 RIR root CA private key theft
1.2.1.2 Intermediate CA exploitation
1.2.1.3 Trust anchor manipulation
1.2.2 Signature validation bypass
1.2.2.1 Algorithm downgrade attacks
1.2.2.2 Weak cryptographic implementation exploitation
1.2.2.3 Signature verification logic flaws
1.2.3 Key management attacks
1.2.3.1 Hardware security module (HSM) vulnerabilities
1.2.3.2 Key generation weaknesses
1.2.3.3 Certificate revocation list manipulation
1.3 Validator targeting [OR]
1.3.1 Relying party software exploitation
1.3.1.1 Buffer overflow attacks
1.3.1.2 Memory corruption vulnerabilities
1.3.1.3 Remote code execution flaws
1.3.2 Cache poisoning attacks
1.3.2.1 Validated ROA payload manipulation
1.3.2.2 Trust chain compromise
1.3.2.3 Stale data persistence exploitation
1.3.3 Protocol implementation attacks
1.3.3.1 RPKI-to-router protocol manipulation
1.3.3.2 Serial number synchronisation attacks
1.3.3.3 Error handling exploitation
1.4 Network infrastructure attacks [OR]
1.4.1 Transport layer targeting
1.4.1.1 RSYNC protocol exploitation
1.4.1.2 HTTPS man-in-the-middle attacks
1.4.1.3 TCP session hijacking
1.4.2 Firewall and ACL bypass
1.4.2.1 Protocol evasion techniques
1.4.2.2 Port and service manipulation
1.4.2.3 Tunnelling attacks
1.4.3 DDoS and resource exhaustion
1.4.3.1 Repository server flooding
1.4.3.2 Validator resource exhaustion
1.4.3.3 Network bandwidth saturation
1.5 Data integrity attacks [OR]
1.5.1 Manifest manipulation
1.5.1.1 Manifest forgery and replay attacks
1.5.1.2 File hash manipulation
1.5.1.3 Manifest expiration exploitation
1.5.2 ROA manipulation
1.5.2.1 Unauthorised ROA creation
1.5.2.2 MaxLength attribute exploitation
1.5.2.3 Origin AS manipulation
1.5.3 Object validation bypass
1.5.3.1 Malformed object processing flaws
1.5.3.2 Parser implementation vulnerabilities
1.5.3.3 ASN.1 decoding exploits
1.6 Configuration and management attacks [OR]
1.6.1 Administrative interface targeting
1.6.1.1 Web portal vulnerabilities
1.6.1.2 API endpoint exploitation
1.6.1.3 Credential theft and brute force attacks
1.6.2 Human factor exploitation
1.6.2.1 Social engineering attacks
1.6.2.2 Configuration error induction
1.6.2.3 Operational procedure manipulation
1.6.3 Backup and recovery targeting
1.6.3.1 Backup system compromise
1.6.3.2 Recovery process manipulation
1.6.3.3 Archive integrity attacks
1.7 Trust chain exploitation [OR]
1.7.1 Trust anchor compromise
1.7.1.1 TAL (trust anchor locator) manipulation
1.7.1.2 Root certificate distribution attacks
1.7.1.3 Trust propagation exploitation
1.7.2 Certificate authority hierarchy attacks
1.7.2.1 Intermediate CA compromise
1.7.2.2 Certificate signing request manipulation
1.7.2.3 Path validation bypass
1.7.3 Revocation mechanism targeting
1.7.3.1 CRL (certificate revocation list) manipulation
1.7.3.2 OCSP stapling attacks
1.7.3.3 Revocation status bypass
1.8 Protocol-specific attacks [OR]
1.8.1 RTR (RPKI-to-router) protocol attacks
1.8.1.1 Session hijacking and manipulation
1.8.1.2 PDU (protocol data unit) forgery
1.8.1.3 Serial number manipulation
1.8.2 RRDP protocol exploitation
1.8.2.1 Delta file manipulation
1.8.2.2 Snapshot integrity attacks
1.8.2.3 Session management vulnerabilities
1.8.3 RSYNC protocol attacks
1.8.3.1 Authentication bypass
1.8.3.2 File transfer manipulation
1.8.3.3 Protocol option exploitation
1.9 Supply chain attacks [OR]
1.9.1 Software distribution compromise
1.9.1.1 Validator software backdoors
1.9.1.2 Update mechanism manipulation
1.9.1.3 Dependency chain exploitation
1.9.2 Hardware compromise
1.9.2.1 HSM (hardware security module) vulnerabilities
1.9.2.2 Network device firmware manipulation
1.9.2.3 Manufacturing process attacks
1.9.3 Third-party service targeting
1.9.3.1 Cloud service compromise
1.9.3.2 CDN infrastructure attacks
1.9.3.3 Hosted RPKI service exploitation
1.10 Advanced persistent threats [OR]
1.10.1 State-sponsored attacks
1.10.1.1 Long-term validator compromise
1.10.1.2 Cryptographic backdoor insertion
1.10.1.3 Infrastructure-wide targeting
1.10.2 Organised crime targeting
1.10.2.1 Ransomware attacks on RPKI infrastructure
1.10.2.2 BGP hijacking for financial gain
1.10.2.3 Data exfiltration for resale
1.10.3 Insider threat exploitation
1.10.3.1 Rogue administrator attacks
1.10.3.2 Privileged credential misuse
1.10.3.3 Policy manipulation
Social engineering, configuration error induction, and operational procedure manipulation (1.6.2) are the human interaction layer of these control paths; the techniques are covered in social engineering of infrastructure operators.
A guard with moving parts¶
The ecosystem is a stack of moving parts, RSYNC, RRDP and RTR, cryptographic operations and distributed repositories, and each part is a surface.
Validators disagree on edge cases and error handling, and the disagreements are exploitable.
Human error is decisive: a wrong MaxLength or a stray AS0 authorisation can invalidate a legitimate route by accident.
Legacy dependence lingers, RSYNC in particular, with its known limitations still in the path.
The trust hierarchy of RIRs, CAs and relying parties has many points that can fail.
Few operators monitor the RPKI infrastructure itself, so an attack on it can run unseen.