Global Surveillance Under the Lens: Reconstructing Passive Dolev-Yao Intruders
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Security Radio. Generated commentary on the latest security and cryptography papers.
Nadia: We're looking at the paper: "Global Surveillance Under the Lens".
Elias: Here's the gist of the paper. This paper introduces an automated, specification-guided methodology designed for reconstructing passive Dolev–Yao intruder executions from captured network traffic while simultaneously validating these reconstructions against a formal protocol specification.
Nadia: The framework operates by taking a PCAP capture, an AnBx specification, and associated deployment metadata as input.
Elias: It executes a multi-stage pipeline: first, it extracts Java-serialised objects from TCP streams while preserving object graphs using TC REFERENCE markers; second, it pairs bidirectional streams into connections and groups them into sessions, enumerating candidate channel assignments when interleaved sessions are present; third, it maps the extracted objects to protocol steps using the compiler-generated channel-to-step map; fourth, it generates a passive intruder role that reads these reconstructed step files and performs the compiler-derived defensive checks; and finally, it scores each candidate arrangement based on the number of clean sessions and passed checks to select the best reconstruction.
Nadia: The methodology specifically addresses three key reconstruction challenges: handling multiple serialised objects and object references within a single TCP stream, resolving ambiguity in mapping interleaved connections to concurrent protocol sessions, and performing semantic validation of candidate reconstructions.
Title and authors: Nadia: So we're looking at this paper titled "Global Surveillance Under the Lens: Reconstructing Passive Dolev-Yao Intruders." It sounds pretty technical, but basically, it tackles how to take raw network traffic and figure out if someone was snooping around a specific communication protocol.
Elias: Yeah, the authors are Paolo Modesti and Teesside University. The title suggests they're trying to reconstruct what a passive attacker did by looking at the network captures themselves, instead of just reading the packets raw.
Nadia: Exactly. They're not just showing you what data is in the packet; they’re trying to figure out if that sequence of messages actually matches something that should have happened according to a formal protocol definition. It moves beyond simple traffic analysis into protocol understanding, which is key for security research because it tells you if the attack succeeded or not.
Priya: From my side, I'm curious about what "passive" really means here. Does this method only work if the attacker is just listening and not actively sending messages back to interfere? That distinction matters when we talk about real-world surveillance scenarios.
Elias: That's a fair point, Priya. The paper focuses on passive Dolev-Yao intruder executions, meaning the intruder just listens and relays what it sees without changing the actual communication flow between the legitimate parties. The goal is to reconstruct that execution path from a PCAP file and an AnBx specification, which is what they call this paper >
The paper's summary: Nadia: Okay, so how does this methodology actually work? The core idea seems to be taking a PCAP file and the protocol specification, and then building something executable—a passive intruder role—that checks if the captured traffic is consistent with what a Dolev-Yao analysis predicts.
Elias: That's right. They have this multi-stage pipeline. First, they extract Java-serialised objects from those TCP streams while keeping track of the object relationships using special markers called TC REFERENCE > then they pair up the bidirectional streams into connections and group them into sessions, figuring out which session corresponds to which protocol step when things are interleaved >
Priya: And I'm wondering about that object extraction part. They're dealing with Java-serialised objects, which sounds complicated for a network capture. What does that extraction process actually look like in practice? Does it handle the complexity of multiple objects being sent in one stream well?
Nadia: It handles it by using those TC REFERENCE markers to preserve the object graphs even when things are packed together > then they map those extracted objects onto the actual protocol steps using a compiler-generated map, which is clever because it uses knowledge from the specification itself >
Elias: And then that generated intruder role performs all these Dolev-Yao checks. Instead of doing symbolic deduction at runtime, which is slow, they materialise that knowledge as receive-side checks in this generated role > essentially making the compiler's symbolic analysis into actual code that runs on the captured data >
Priya: So the paper's big claim here seems to be that it connects the formal specification directly to concrete network traffic. That means if a trace fails these checks, we know for sure it wasn't what was supposed to happen according to the protocol rules >
The paper's improvements: Nadia: Now, the authors talk about what they improved on previous work or what their approach does differently. They focus on solving three main problems: handling multiple serialised objects and object references in a single TCP stream > resolving ambiguity when connections are interleaved into concurrent protocol sessions > and performing that semantic validation of the candidate reconstructions we talked about earlier >
Elias: The key improvement, as they frame it, is creating a concrete operational instantiation of this passive intruder where the compiler's symbolic knowledge analysis is materialized as executable receive-side checks in the generated role > that's making sure you get actual code back to test against >
Priya: From a measurement standpoint, I'm interested in how they handle the ambiguity of interleaved connections. When you have multiple sessions happening at once, distinguishing which session belongs to which protocol step is tricky because of timing issues and how the streams are ordered >
Nadia: They tackle that by enumerating candidate channel assignments when those interleaved sessions exist > and then they score each candidate arrangement based on a metric, picking the best one with fewer clean sessions and passed checks > it’s a guided selection process rather than just picking the first plausible thing they find >
Elias: And that scoring is done using Dolev-Yaoderived consistency checks to distinguish between competing reconstructions of the same traffic, which helps them select the most likely true execution path > it's about formal consistency guiding the reconstruction, not just heuristics >
Conclusion: Nadia: So to wrap up, this paper presents a framework that automates reconstructing passive Dolev-Yao intruder executions by extracting serialised objects and validating them against a formal protocol specification using those compiler-derived checks.
Elias: The main results show strong performance across different scenarios. For single session captures, they got ninety-four point two percent successful reconstruction of the expected objects > and for protocols that can handle parallel connections, they achieved eighty-nine point three percent successful reconstruction for both two and three concurrent sessions while recovering about ninety-seven point two percent of the expected objects >
Priya: Those numbers suggest that this method is quite robust when dealing with different protocol complexities, but I'm still watching how those checks handle actual subtle mutations in the data. They mention that they can flag check failures or check-count deficits to tell if a change hit content-dependent checks versus just structural properties like the step count >
Nadia: Exactly. It gives us a tool where we can test if an observed trace is actually consistent with the specification, not just structurally sound on its own > it's a proof of concept for how we can instantiate passive adversary models in practice >
Elias: It shows that the passive adversary model can be concrete, and that's important because it links formal protocol specifications to real network traffic in a tangible way >
Priya: It’s a solid step forward for automated surveillance analysis, showing what this kind of toolchain could actually do in a real security monitoring environment >
Nadia: Yeah, the paper "Global Surveillance Under the Lens: Reconstructing Passive Dolev-Yao Intruders" gives us a concrete way to test if traffic matches protocol expectations. We'll keep an eye on how they apply these reconstruction techniques to even more complex systems next >
Paolo Modesti
Teesside University
cs.CR, cs.SE
Submitted: 2026-11-04
Updated: 2026-11-04
Comments: 50 pages, 7 figures
Code: https://github.com/mark4865/JavaCryptoClientServer
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
Importance score: 83/100
The gist: This paper introduces an automated, specification-guided methodology designed for reconstructing passive Dolev–Yao intruder executions from captured network traffic while simultaneously validating…
Key concepts
- Java-serialised objects
- These are data structures serialized into a format that allows them to be transmitted over TCP streams. The framework extracts these objects from network captures while preserving their internal structure using special markers, which is crucial for understanding the actual data being exchanged during the intrusion.
- AnBx specification
- This is a formal, mathematical description of a communication protocol. It acts as the ground truth against which captured network traffic is checked. The system uses this specification to generate defensive checks that verify if an intruder's actions are consistent with the defined protocol rules.
- Dolev-Yao intruder model
- This is a theoretical model of an adversary who can intercept, modify, and replay messages in a network. The framework attempts to reconstruct the specific sequence of actions this abstract intruder took by analyzing the captured traffic.
- Specification-guided candidate selection
- When multiple ways to interpret the captured traffic exist, this technique uses consistency checks derived from the formal specification. It scores different possible reconstructions, favoring arrangements that have more 'clean' sessions and pass all required protocol validations.
Terminology
Summary
This paper introduces an automated, specification-guided methodology designed for reconstructing passive Dolev–Yao intruder executions from captured network traffic while simultaneously validating these reconstructions against a formal protocol specification. p. 0 Global Surveillance Under the Lens:
Reconstructing Passive Dolev–Yao Intruders
Paolo Modesti
Teesside University, Middlesbrough, United Kingdom
Abstract
Global surveillance systems record large volumes of network traffic for retrospective analysis, but packet captures alone reveal little about protocol-level
semantics. This paper presents an automated, specification-guided methodology for reconstructing passive Dolev–Yao intruder executions from captured
network traffic and validating the reconstructed executions against a formal protocol specification. This paper presents an automated, specification-guided methodology for reconstructing passive Dolev–Yao intruder executions from captured
network traffic and validating the reconstructed executions against a formal protocol specification. The framework takes a PCAP capture, an AnBx specification, and the associated deployment metadata; it extracts Java-serialised objects
from TCP streams, reconstructs protocol sessions and channel-to-step assignments, and generates a passive intruder that evaluates the Dolev–Yao-derived
checks produced by the AnBx compiler.Where the paper says it
Where the paper says it
The methodology specifically addresses three key reconstruction challenges: handling multiple serialised objects and object references within a single TCP stream, resolving ambiguity in mapping interleaved connections to concurrent protocol sessions, and performing semantic validation of candidate reconstructions. p. 0 Three reconstruction challenges are addressed: multiple serialised objects and object references within a TCP stream,
ambiguity in mapping interleaved connections to concurrent protocol sessions,
and semantic validation of candidate reconstructions. We evaluate the methodology on 121 protocols from the AnBx test suite. The contributions of this paper are fourfold. First, a concrete operational
instantiation of a passive Dolev–Yao intruder in which the compiler’s symbolic
knowledge analysis is materialised as executable receive-side checks over reconstructed observations. Third, specification-guided candidate selection, in which Dolev–Yaoderived consistency checks distinguish competing reconstructions of the same
captured traffic. Fourth, an evaluation over 121 protocols assessing behaviour
under multiple sessions, alternative cryptographic providers, third-party Java
traffic, mismatched protocol-role mappings, and deliberately mutated traces.Where the paper says it
Where the paper says it
Where the paper says it
The framework is rigorously evaluated across 121 protocols from the AnBx test suite. p. 2 The intruder does
not modify messages; it observes, validates, and reconstructs them, faithfully
modelling a passive adversary. We evaluate the framework on a test suite of 121 protocols drawn from
the OFMC distribution [5] that ships with the AnBx toolchain and from the
SET [15] and iKP [16] families. For single-session captures,
the standard JDK cryptographic provider recovers 97. 8% of expected objects
and achieves 94. 8% of expected objects
and achieves 94. 2% successful reconstruction, where success requires complete
object recovery and a clean passive-intruder execution. 2% successful reconstruction, where success requires complete
object recovery and a clean passive-intruder execution. For the 112 parallelcapable protocols, the toolchain achieves 89. For the 112 parallelcapable protocols, the toolchain achieves 89. 3% successful reconstruction for
both two and three concurrent sessions, while recovering approximately 97. For single-session captures,
the standard JDK cryptographic provider recovers 97. 8% of expected objects
and achieves 94. The contribution is not that a passive adversary can infer all
of this from an arbitrary PCAP; it is that, once this knowledge is available, the
semantic reconstruction itself can be automated. The extraction stage does not
depend on the compiler having produced the implementation, and the defensive
checks are derived from the specification, not from the implementation; both
claims are substantiated in Sections 4. The purpose of the surveillance framing is defensive: the framework makes explicit what a passive adversary can recover from captured traffic, so that the
resulting knowledge can inform the design of protocols that are more resilient to
retrospective analysis. The prototype does not model a complete intelligencecollection system or infer unknown protocols from arbitrary traffic; it provides
a controlled laboratory model of one stage of retrospective surveillance analysis,
and a basis for assessing which protocol design choices reduce the information
recoverable after interception.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Automated passive Dolev–Yao intruder instantiation: The framework can reconstruct passive Dolev–Yao intruder executions from captured network traffic by extracting Java-serialised objects and validating them against a formal protocol specification using generated receive-side checks.
-
Specification-guided candidate selection: The system uses Dolev–Yaoderived consistency checks to distinguish competing reconstructions of the same captured traffic, selecting the best arrangement based on a lexicographic tuple of clean sessions and failed checks.
-
Detection of semantic inconsistencies: The framework detects modifications to valid reconstructions by flagging check failures or check-count deficits, allowing it to identify whether a mutation affects content-dependent checks versus structural properties like step count.
Abstract
Global surveillance systems record large volumes of network traffic for retrospective analysis, but packet captures alone reveal little about protocol-level semantics. This paper presents an automated, specification-guided methodology for reconstructing passive Dolev-Yao intruder executions from captured network traffic and validating the reconstructed executions against a formal protocol specification. The framework takes a PCAP capture, an AnBx specification, and the associated deployment metadata; it extracts Java-serialised objects from TCP streams, reconstructs protocol sessions and channel-to-step assignments, and generates a passive intruder that evaluates the Dolev-Yao derived checks produced by the AnBx compiler. Three reconstruction challenges are addressed: multiple serialised objects and object references within a TCP stream, ambiguity in mapping interleaved connections to concurrent protocol sessions, and semantic validation of candidate reconstructions. We evaluate the methodology on 121 protocols from the AnBx test suite. For single-session captures, the standard JDK cryptographic provider recovers 97.8% of expected objects and achieves 94.2% successful reconstruction, where success requires complete object recovery and a clean passive-intruder execution. For the 112 parallel-capable protocols, the toolchain achieves 89.3% successful reconstruction for both two and three concurrent sessions, while recovering approximately 97.2% of expected objects. Experiments with mismatched protocol-role mappings and deliberately mutated traces show that the generated Dolev-Yao checks respond to semantic inconsistencies. By making the reconstruction step explicit, the solution produces protocol-level observations on which assessments of protocol resilience and the design of defensive countermeasures can be built.
Sources
- On the Practical Feasibility of Harvest-Now, Decrypt-Later Attacks
- Molly: A Verified Compiler for Cryptoprotocol Roles
Related papers
- From Strategy Search to Policy Learning: MCTS-Guided Adversarial Co-Evolution for LLM Safety Alignment
- Privacy Leakage from Recurrent State in LLM Inference
- Incrementally Verifiable Oracle Aided Computation
- Privilege and confidentiality in generative AI workflows
- BadMoE: Backdooring Mixture-of-Experts LLMs via Routing-Aware Triggers and Task-Underutilized Experts
- CoVault: Secure, Scalable Analytics of Personal Data