Patience, key compromise, and staying wrong quietly¶

Attack pattern¶

Everything else in this family assumes the attack has an event to it. This branch assumes the opposite. A key made weak at generation, through a thin entropy source or a backdoored accelerator, produces signatures that validate perfectly for as long as the key lives, and nothing in the routing system is looking for a correct signature from the wrong key. A trust anchor moved gradually, a validation policy edited one line per quarter, a signature check corrupted rather than disabled: each is shaped to survive review instead of defeating it. What distinguishes the branch is duration rather than sophistication. A check that fails occasionally reads as a bug and gets a ticket, where a check that is switched off reads as an incident and gets a response.

1. Advanced Persistent Threat Techniques [OR]

    1.1 Long-Term Key Compromise [OR]

        1.1.1 Supply Chain Attacks [OR]
            1.1.1.1 Hardware backdoors in crypto accelerators
            1.1.1.2 Compromised software distributions
            1.1.1.3 Malicious contributor code injections

        1.1.2 Key Generation Weaknesses [OR]
            1.1.2.1 Weak entropy sources during key generation
            1.1.2.2 Algorithm-specific bias introduction
            1.1.2.3 Compromised random number generators

        1.1.3 Key Storage Compromise [OR]
            1.1.3.1 Cold storage extraction techniques
            1.1.3.2 Cloud HSM configuration breaches
            1.1.3.3 Multi-party computation failures

    1.2 Stealthy Validation Manipulation [OR]

        1.2.1 Low-and-Slow Attacks [OR]
            1.2.1.1 Subtle signature validation corruption
            1.2.1.2 Gradual trust anchor manipulation
            1.2.1.3 Incremental policy modification

        1.2.2 False Flag Operations [OR]
            1.2.2.1 Attribution obfuscation through intermediate ASes
            1.2.2.2 Victim fingerprint spoofing
            1.2.2.3 Third-party tool exploitation

        1.2.3 Persistence Mechanisms [OR]
            1.2.3.1 Reinfection capabilities
            1.2.3.2 Multiple compromise vectors
            1.2.3.3 Anti-forensic techniques

    1.3 AI-Enhanced BGPsec Attacks [OR]

        1.3.1 Machine Learning Exploitation [OR]
            1.3.1.1 AI-generated optimal attack timing
            1.3.1.2 Neural network-based evasion patterns
            1.3.1.3 Reinforcement learning for policy exploitation

        1.3.2 Automated Vulnerability Discovery [OR]
            1.3.2.1 AI-assisted fuzz testing for BGPsec
            1.3.2.2 Machine learning for side-channel detection
            1.3.2.3 Automated exploit generation

        1.3.3 Adaptive Attack Systems [OR]
            1.3.3.1 Self-modifying attack code
            1.3.3.2 Dynamic protocol manipulation
            1.3.3.3 Intelligent countermeasure evasion

Correct signatures from the wrong key¶

A key weak at generation is indistinguishable from a strong one afterwards. Thin entropy and biased generators produce certificates that validate cleanly for their whole lifetime.
Supply chain reaches back further than any protocol control. A compromised accelerator or a malicious commit predates every check the routing system runs.
Gradual change survives review. A trust anchor moved slowly and a policy edited a line at a time each read as maintenance in a diff.
Corrupting validation quietly beats disabling it, because an occasional failure attracts a ticket and an absent check attracts an investigation.
Intermediate autonomous systems supply attribution cover, which is the false flag problem arriving in routing.
Duration is what the branch buys. None of it is faster than the alternatives, and none of it needs to be.