Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LAPRAD: LLM-Assisted PRotocol Attack Discovery

R. Can Aygun, Yehuda Afek, Anat Bremler-Barr, Leonard Kleinrock

TL;DR

Paper tackles improving protocol security by automating vulnerability discovery with LLMs. The authors propose LAPRAD, a three-stage workflow combining idea generation with GPT-o1, attack configuration via ReACT/LangChain, and validation, applied to DNS and DNSSEC. They demonstrate three novel SigCacheFlush variants that degrade resolver capacity and also re-discover existing attacks such as KeyTrap and NRDelegation, validating the approach in Docker and cloud experiments. The work suggests a general, automatable framework for LLM-guided vulnerability research with potential applicability to other well-documented protocols.

Abstract

With the goal of improving the security of Internet protocols, we seek faster, semi-automatic methods to discover new vulnerabilities in protocols such as DNS, BGP, and others. To this end, we introduce the LLM-Assisted Protocol Attack Discovery (LAPRAD) methodology, enabling security researchers with some DNS knowledge to efficiently uncover vulnerabilities that would otherwise be hard to detect. LAPRAD follows a three-stage process. In the first, we consult an LLM (GPT-o1) that has been trained on a broad corpus of DNS-related sources and previous DDoS attacks to identify potential exploits. In the second stage, a different LLM automatically constructs the corresponding attack configurations using the ReACT approach implemented via LangChain (DNS zone file generation). Finally, in the third stage, we validate the attack's functionality and effectiveness. Using LAPRAD, we uncovered three new DDoS attacks on the DNS protocol and rediscovered two recently reported ones that were not included in the LLM's training data. The first new attack employs a bait-and-switch technique to trick resolvers into caching large, bogus DNSSEC RRSIGs, reducing their serving capacity to as little as 6%. The second exploits large DNSSEC encryption algorithms (RSA-4096) with multiple keys, thereby bypassing a recently implemented default RRSet limit. The third leverages ANY-type responses to produce a similar effect. These variations of a cache-flushing DDoS attack, called SigCacheFlush, circumvent existing patches, severely degrade resolver query capacity, and impact the latest versions of major DNS resolver implementations.

LAPRAD: LLM-Assisted PRotocol Attack Discovery

TL;DR

Paper tackles improving protocol security by automating vulnerability discovery with LLMs. The authors propose LAPRAD, a three-stage workflow combining idea generation with GPT-o1, attack configuration via ReACT/LangChain, and validation, applied to DNS and DNSSEC. They demonstrate three novel SigCacheFlush variants that degrade resolver capacity and also re-discover existing attacks such as KeyTrap and NRDelegation, validating the approach in Docker and cloud experiments. The work suggests a general, automatable framework for LLM-guided vulnerability research with potential applicability to other well-documented protocols.

Abstract

With the goal of improving the security of Internet protocols, we seek faster, semi-automatic methods to discover new vulnerabilities in protocols such as DNS, BGP, and others. To this end, we introduce the LLM-Assisted Protocol Attack Discovery (LAPRAD) methodology, enabling security researchers with some DNS knowledge to efficiently uncover vulnerabilities that would otherwise be hard to detect. LAPRAD follows a three-stage process. In the first, we consult an LLM (GPT-o1) that has been trained on a broad corpus of DNS-related sources and previous DDoS attacks to identify potential exploits. In the second stage, a different LLM automatically constructs the corresponding attack configurations using the ReACT approach implemented via LangChain (DNS zone file generation). Finally, in the third stage, we validate the attack's functionality and effectiveness. Using LAPRAD, we uncovered three new DDoS attacks on the DNS protocol and rediscovered two recently reported ones that were not included in the LLM's training data. The first new attack employs a bait-and-switch technique to trick resolvers into caching large, bogus DNSSEC RRSIGs, reducing their serving capacity to as little as 6%. The second exploits large DNSSEC encryption algorithms (RSA-4096) with multiple keys, thereby bypassing a recently implemented default RRSet limit. The third leverages ANY-type responses to produce a similar effect. These variations of a cache-flushing DDoS attack, called SigCacheFlush, circumvent existing patches, severely degrade resolver query capacity, and impact the latest versions of major DNS resolver implementations.
Paper Structure (14 sections, 4 figures, 1 table)

This paper contains 14 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. Threat Model
  4. LAPRAD methodology for DNS
  5. LLM-Assisted Protocol Attack Discovery (LAPRAD)
  6. Discovering New DNS Vulnerabilities
  7. Step-1: Attack Idea Investigation
  8. Step-2: Attack Configuration Generation
  9. Source of Information
  10. Proposed Mitigations
  11. Re-discovering Existing DNS Vulnerabilities
  12. Experiments
  13. Discussions and Future Work
  14. Conclusion

Figures (4)

  • Figure 1: Attack Investigation Workflow. The human figure indicates where human assistance is required to complete the task. The dashed line connects Output-1 (the attack idea) from Step-1 to Step-3, illustrating that it will be tested using the attack zone configuration (Output-2).
  • Figure 2: Bait & Switch Cache Flushing Attack
  • Figure 3: For the top 9,336 DNSSEC-enabled domains and their TLDs, panel (a) presents the DNSSEC response packet size distribution, while panel (b) shows the DNSKEY distribution (up to 16 keys per domain, with values from 11 to 16 keys omitted as they account for only 0.05% of domains).
  • Figure 4: a)Resolver Benign Max QPS Performance with Bait&Switch Attack (excluding SERVERFAIL responses) under constant attack rates of 0, 300, 1,000, and 3,000 QPS. b) Detailed benign response rates of resolvers benign max QPS under a bait-and-switch attack at constant 1,000 QPS. The attack starts at 0s, marked by the black arrow.