Authenticated Browsing: Reconciling Application-Layer Flooding Defense with Data Confidentiality
Year: 2026
Type: article
Abstract: Organizations in regulated sectors, including financial institutions subject to banking secrecy, healthcare providers handling patient data, and government services processing citizen information, face a structural obstacle in deploying effective denial-of-service (DoS) defenses against application-layer flooding: the most capable commercial solutions require third-party TLS termination, which conflicts with their data protection obligations. We systematically analyze existing DoS mitigation approaches, both industry-deployed and academic, and show that none simultaneously achieve application-layer traffic differentiation and compatibility with stringent confidentiality requirements at line rate, without per-connection state or network changes. We present Authenticated Browsing, a system that separates traffic differentiation from traffic inspection by making prior application-layer authentication verifiable at the packet level. Authenticated Browsing repurposes widely deployed identity infrastructure for DoS mitigation: after a user is verified as enrolled with the protected service, an Identity Provider (IdP) issues short-lived cryptographic credentials from which a user agent derives per-packet authentication tags. Forward-deployed validators verify these tags at line rate using only symmetric cryptography, without inspecting payload content or learning user identities. The IdP's role is limited to credential issuance and need not coincide with the application's login system. We implement Authenticated Browsing and evaluate it at scale. A single validator sustains more than 70 million authenticated packets per second on commodity hardware, achieving line rates of 160 Gbps. Internet measurements show that Authenticated Browsing authentication headers are preserved across a substantial fraction of real-world paths. These results show that application-layer flooding attacks can be mitigated under strong confidentiality requirements.
DOI: 10.1145/3856957
Institutions: ETH Zurich, University of Amsterdam
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7236844540
