Internet-wide scanning is crucial for security researchers and practitioners. Exhaustively mapping the entire IPv4 address space has facilitated the development of new cybersecurity protocols, the identification and tracking of vulnerabilities, the improvement of DDoS defenses, and the exposure of global Internet censorship. IPv6 addresses are 128 bits long, with every consecutive 4 bits having a hexadecimal representation called a nybble. The vast address space of IPv6—340 trillion trillion trillion addresses—precludes exhaustive scanning, necessitating the development of new methods specifically designed for IPv6. While fast Internet scanning has helped answer numerous questions in Internet security, this understanding is limited to the 32-bit IPv4 address space. Current tools employ an exhaustive scanning methodology, sending packets to every single IPv4 address. However, this brute force approach does not scale to IPv6’s approximately 340 trillion trillion trillion addresses. Tasks: Leverage IPv6 address and autonomous system (AS) deployment patterns during address generation. Sample known IPv6 allocations using a weighted Reinforcement Learning update technique. Send these variable-length allocations to an LSTM model to generate the upper 64 bits of addresses. Use an iterative address generator that employs domain-specific knowledge to pair the upper 64 bits with likely active lower 64 bits. Generate the lower 64 bits based on observed patterns.

Internet-wide scanning is crucial for security researchers and practitioners. Exhaustively mapping the entire IPv4 address space has facilitated the development of new cybersecurity protocols, the identification and tracking of vulnerabilities, the improvement of DDoS defenses, and the exposure of global Internet censorship. IPv6 addresses are 128 bits long, with every consecutive 4 bits having a hexadecimal representation called a nybble. The vast address space of IPv6—340 trillion trillion trillion addresses—precludes exhaustive scanning, necessitating the development of new methods specifically designed for IPv6. While fast Internet scanning has helped answer numerous questions in Internet security, this understanding is limited to the 32-bit IPv4 address space. Current tools employ an exhaustive scanning methodology, sending packets to every single IPv4 address. However, this brute force approach does not scale to IPv6’s approximately 340 trillion trillion trillion addresses. Tasks: Leverage IPv6 address and autonomous system (AS) deployment patterns during address generation. Sample known IPv6 allocations using a weighted Reinforcement Learning update technique. Send these variable-length allocations to an LSTM model to generate the upper 64 bits of addresses. Use an iterative address generator that employs domain-specific knowledge to pair the upper 64 bits with likely active lower 64 bits. Generate the lower 64 bits based on observed patterns.