RPKI Now Covers 70 Percent of Internet Routes, But Path Hijacks Still Get Through

Roughly 70.3% of the internet's routed prefixes now carry a valid Route Origin Authorization (ROA), according to IPinfo's September 2026 measurement — up from 54.0% a year earlier. That is the fastest 12-month jump in RPKI adoption since the protocol was standardized, and it means a majority of the global routing table is now protected against the simplest form of BGP hijacking: an attacker announcing a prefix they have no right to originate.
The catch is in what that number actually measures. Route Origin Validation (ROV) checks exactly one thing — whether the autonomous system announcing a prefix is authorized to do so. It says nothing about the path that announcement takes to get there, and the mechanism built to close that gap, ASPA (Autonomous System Provider Authorization), is deployed on roughly 2.5% of ASNs according to Hurricane Electric's July 2026 figures. The industry solved the easy half of BGP security and is now stalled at the hard half.
Why BGP Was Never Designed to Be Trusted
Border Gateway Protocol dates to 1989 and operates on a simple premise: every network tells its neighbors which address blocks it can reach, and every neighbor believes it. There is no cryptographic proof baked into the protocol itself — just reputation, manual filtering, and, for decades, hope. A misconfigured router or a malicious operator could announce someone else's address space and a meaningful fraction of global traffic would simply route there, no exploit required.
RPKI, finalized as an IETF standard in the early 2010s, attaches a cryptographically signed certificate to each address block, specifying which AS is allowed to originate it. Routers that perform ROV reject announcements that don't match. The deployment problem was never technical — it was coordination. ROV only protects you if the networks between you and an attacker also filter on it, and for over a decade, too few did for anyone's investment to pay off. 2026 is the year that calculus flipped.
What Actually Changed
Three numbers explain the shift. ARIN's September 30, 2026 snapshot puts RPKI-valid coverage of ARIN-registered IPv4 space at 59.4%. Hurricane Electric's July 2026 data shows IPv4 ROA coverage at 65.6% and IPv6 at 73.5% — IPv6 networks, being newer and run by operators who set up RPKI from day one, are consistently ahead. And in February 2026, Tier-1 transit provider Sparkle began outright rejecting RPKI-invalid route announcements rather than just logging them, a move that network engineers have been predicting for years as the point where RPKI adoption stops being optional. When a Tier-1 drops your invalid routes, your traffic simply doesn't reach parts of the internet — that's a much stronger incentive than a security audit finding.
The practical result shows up in incident counts. Qrator's Q1 2026 global routing security report counted one global BGP hijack and seven global route leaks for the quarter — a small number by historical standards, though route leaks (where legitimate paths get announced in the wrong direction, often by misconfiguration rather than malice) are exactly the category ROV does nothing to stop.
The Blind Spots Nobody's Dashboard Shows
ROV has at least two structural gaps that a 70% coverage number obscures. First, many ROAs are registered with a loose max-length attribute — meaning the certificate authorizes not just the exact prefix but any more-specific sub-prefix within it. An attacker can announce a more specific route for a smaller slice of address space, have it validate as RPKI-correct, and still win the routing contest against the legitimate, less-specific announcement, because more-specific routes are preferred by default in BGP. This is a deaggregation attack, and tightening max-length values to the minimum necessary is a config change operators routinely skip because it's tedious and rarely breaks anything until it does.
Second, ROV validates the origin, not the path. An attacker who compromises or colludes with a legitimately-authorized AS can still forge or manipulate the path segments after that point, and ROV has no mechanism to catch it. ASPA closes this gap by letting an AS cryptographically declare its legitimate upstream providers, so a path through an unauthorized provider gets flagged — but at 2.5% deployment, it's nowhere near providing network effects yet.
What Network Operators Should Actually Do
Three concrete actions separate real protection from a checkbox. First, audit existing ROAs for overly permissive max-length values — a ROA for a /24 shouldn't authorize max-length /24 unless sub-allocation is actually happening; tightening it closes the deaggregation hole immediately. Second, deploy ASPA now even though adoption is low — the protocol only helps once enough of the path participates, and someone has to be early. Third, don't treat RPKI as a complete routing security program; pair it with real-time BGP monitoring (Qrator, Cloudflare Radar, and RIPE's RIS all publish usable feeds) that catches route leaks and anomalies ROV structurally cannot see. The 70% milestone is real progress. It is not the finish line.