Things that Matter -- Identifying Interactions and IoT Device Types in Encrypted Matter Traffic
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: "Things that Matter -- Identifying Interactions and IoT Device Types in Encrypted Matter Traffic".
Elias: Here's the gist of the paper. Matter is designed to enable interoperability across IoT products of various manufacturers, independently of underlying communication technologies.
Nadia: However, passive eavesdroppers can infer interactions and device types by analyzing encrypted Matter traffic.
Elias: Research Question and Objectives The main research question is what information a fully passive attacker can infer by analyzing encrypted Matter traffic.
Nadia: This overarching question is broken down into two sub-questions: RQ1 investigates how accurately a passive attacker can identify specific packets, transactions, and types of interactions from the analysis of encrypted Matter traffic.
Title and authors: Nadia: So we're talking about this paper now, "Things that Matter -- Identifying Interactions and IoT Device Types in Encrypted Matter Traffic." It’s really about how much information you can actually get just by listening to encrypted traffic from these new smart devices.
Elias: Yeah, it tackles the fact that Matter is supposed to be interoperable across different companies, but this design might actually make it easier for someone passively listening to the traffic to figure out what kind of device they are dealing with and what's happening.
Priya: From a privacy standpoint, that’s exactly what we need to look at because if you can tell the difference between a light bulb and a lock just by looking at the data, that compromises user anonymity across all those devices.
Nadia: Right, so the main idea is figuring out what specific things an attacker can actually pull out of that encrypted stream. The paper sets up two main questions to guide their study: how accurately can someone identify individual packets and transactions, and how well can they figure out the specific device types involved.
Elias: That’s a good way to frame it because those two sub-questions are what actually drive the whole analysis in this work. They want to know the limits of inference from encrypted traffic.
Priya: I'm curious about how they handle all that simulated and real-world data, because just having a theoretical model isn't enough; we need to see if it holds up in messy situations.
Nadia: Well, they use a lot of data, including five million plus real-world Matter packets collected over four weeks in their dataset D0, which serves as this baseline to look for patterns involving specific device types. They also use datasets D1 and D2 to test how packet loss and delays affect how well they can classify things.
Elias: That setup is interesting because the authors aren't just looking at a clean stream; they’re testing the robustness of their identification methods against real-world noise like packet loss or those timing variations.
Priya: And when we look at the results for identifying individual interactions, what’s actually happening there? Does it manage to distinguish between different types of commands?
Nadia: They found patterns that let them identify various interaction types in the encrypted Matter traffic, including things like multi-attribute Read Requests and single-attribute Read Requests, empty Read Data Actions, non-empty Read Data Actions, End sequences, Write Request Action packets like WRA one and WRA two packets as well as IRA one and IRA two interactions.
Title and authors: Elias: That’s a lot of specific transaction types they were able to spot in the traffic analysis. They claim they achieved accuracy of ninety-nine point eight seven percent for identifying all those patterns, with a recall of ninety-nine point five zero percent and a precision of ninety-nine point nine six percent and an F1 score of ninety-nine point seven two percent.
Priya: Ninety-eight percent accuracy on transaction identification sounds strong, but I need to know how reliable that is when the conditions get slightly worse, like with some packet loss or delay?
Nadia: They said these results were pretty solid even when there was low levels of packet loss and delays, performing comparably to regular conditions with an average packet loss of five percent and ten percent. It seems they're quite resilient to those kinds of common network hiccups.
Elias: Moving over to the second part, which is about classifying the device types themselves. The paper establishes these fingerprints based on specific things like the lengths of IRA one packets and the sequence of IRA one and TRA packets as distinct markers for different device categories.
Priya: So, they’re not just looking at what's happening in a transaction, but they’re building a fingerprint for the device itself—like separating lighting appliances from lock appliances or sensors?
Nadia: Exactly. They found these distinct fingerprints revealing categories like lighting appliances, lock appliances, plug appliances and sensors. When they tested this on dataset D3, the methodology achieved one hundred percent for all four performance metrics in identifying those specific device types.
Elias: That one hundred percent score is impressive if it’s true across the board. But the authors also flagged a limitation: performance drops when command diversity is low, which suggests that if you see a lot of commands going to the same device, it becomes harder for their system to tell them apart.
Priya: So, what does that mean in practice? If someone uses a simple device with repetitive commands, the attacker has more trouble classifying it correctly?
Nadia: It means that if you have low command diversity, the attacker's ability to accurately classify the device type decreases. They found this when observing many commands issued to one specific device.
Elias: Overall, what does this all mean for us listening right now? The paper points out that even though Matter is encrypted at the application layer, things like Read, Write, Invoke and Report transactions remain identifiable through consistent packet lengths generated by Matter's design.
Priya: So the implication is that the very effort to make Matter interoperable across different vendors inadvertently creates a leak where an eavesdropper can still determine device type and interaction even with strong encryption in place.
Title and authors: Nadia: That’s the core concern, Priya. The finding implies that Matter's focus on interoperability introduces a privacy leakage affecting all devices collectively, not just one specific implementation.
Elias: The authors suggest that mitigations could involve introducing controlled and synchronized variability into the packet structures, like randomized packet padding, which would help protect against traffic analysis while still keeping the interoperability working.
Priya: So it’s a trade-off: you want device interoperability but you also need to add noise or randomness to the traffic structure to hide those patterns from passive analysis.
Nadia: Right, so we’re seeing that without these countermeasures against traffic analysis, next-generation protocols like Matter risk undermining the privacy they were designed to protect. The consistent patterns in encrypted traffic generated by Matter-enabled IoT devices allow for the inference of transactions and device types.
Elias: It really raises significant concerns about user privacy in any environment that relies on Matter for connecting our smart things. We’re seeing how these design choices create tangible security risks when analyzed passively, even with strong encryption applied at the application layer.
Priya: I think the focus moving forward has to be on ensuring that these necessary interoperability features don't become pathways for traffic analysis attacks. The data shows we have a clear way to classify things, so now we need to build defenses against that classification method.
Nadia: Exactly, so the paper "Things that Matter -- Identifying Interactions and IoT Device Types in Encrypted Matter Traffic" shows us exactly what an attacker can do when they just listen to the encrypted stuff. It gives us concrete examples of how transaction types and device categories are exposed through packet structures.
Elias: It’s a good technical piece because it moves beyond just saying "encryption is good" and shows precisely where the structure of the communication itself leaves holes that passive analysis can exploit. That level of detail is crucial for anyone trying to secure these systems properly.
Priya: I agree, it’s less about whether encryption works and more about what information leaks through its shape. Understanding those packet lengths is key to building better privacy controls at the protocol level, not just on top of it.
Nadia: So that's where we leave things for this paper. It shows that we need controlled variability in packet structures to protect against this kind of traffic analysis while still keeping the interoperability working across all those devices.
Elias: And moving on, we’ve got a lot more on these topics, so keep an eye out for our next look at something different.
The paper's summary: Nadia: So, to recap, this paper looks at how much information an attacker can pull out just by listening to encrypted Matter traffic.
Elias: Right, they break it down into two main questions: first, how well can you identify individual packets and transactions? And second, how accurately can you figure out what kind of device is talking over the air?
Nadia: Exactly. The researchers found that with this method, they can classify specific things like Read requests and Write requests in encrypted traffic with nearly ninety-nine percent accuracy.
Elias: That's pretty high for passive analysis, considering the encryption at the application layer. They established distinct fingerprints based on packet lengths and sequences for different device types—things like lighting appliances or locks.
Nadia: But there's a catch; their results showed that if you have low command diversity, meaning many commands going to one device are all similar, the attacker has a harder time classifying that specific device type accurately.
Elias: That suggests the structure of the communication itself leaves these patterns exposed even when you’re using strong encryption. The big implication here is that Matter's focus on making things work together across different companies introduces this sort of collective privacy leak.
Nadia: It means that just because a connection is encrypted doesn't mean it's completely hidden from someone who can analyze the shape of the data coming across the wire.
Elias: So, they propose a way to counter this by introducing controlled and synchronized variability into how those packets are structured, maybe some randomized padding.
Nadia: It’s a trade-off, I guess; you want interoperability but you also need to add noise to hide the patterns that could be used for traffic analysis.
Elias: And that leads us to thinking about how we can actually build defenses against these kinds of structural leaks in the future.
The paper's improvements: Nadia: So, we're talking about how they suggest making these systems more resilient against traffic analysis attacks in this paper.
Elias: They found that to protect privacy without killing interoperability, you need to introduce controlled and synchronized variability into the packet structures itself.
Nadia: Right, they're talking about something like randomized packet padding, which adds noise to the communication pattern so it’s harder for an eavesdropper to find those consistent fingerprints.
Elias: That means if every read request always has the exact same length or sequence, that’s a target. Introducing variation breaks that consistency.
Nadia: It’s about making sure the shape of the traffic isn't a perfect fingerprint for a specific interaction or device type.
Elias: They argue this is essential because right now, without it, even application layer encryption doesn't stop passive analysis when you look at packet lengths and sequences.
Nadia: So the suggestion is that developers need to build in this structural randomness from the start to protect user privacy across all those Matter devices.
Elias: It's a design consideration, shifting the focus from just keeping the application data encrypted to also making sure the communication pattern itself is noisy enough.
Nadia: That makes sense for practical defense because it addresses that specific weakness we saw—the predictable packet lengths and sequences.
Elias: And if they get this right, it helps maintain that interoperability you're worried about, because the variability would be controlled and synchronized across different devices.
Conclusion: Tom: So to wrap up, we've seen how analyzing encrypted Matter traffic lets an AI infer specific transactions and device types, even though they use strong encryption at the application layer.
Nadia: It really shows that the design choices made for interoperability in Matter inadvertently create a leakage where passive listeners can still figure out what's going on.
Elias: That's the core issue; because of how Matter is built, consistent packet lengths and sequences act like fingerprints for specific device types and interactions.
Priya: I’m just wondering what this means for people who are relying on these smart home systems to be private? If an AI can tell if you have a lock or a light just by listening, that changes the whole picture.
Nadia: It does, Priya. The paper "Things that Matter -- Identifying Interactions and IoT Device Types in Encrypted Matter Traffic" highlights how this structural pattern recognition works in practice with specific results on transaction identification accuracy.
Elias: Exactly. They show that the consistency of Read and Write requests is predictable enough to allow for high-accuracy classification when they look at the measurement record over time.
Priya: But you mentioned that performance drops when command diversity is low, which tells us there are scenarios where this attack becomes much more effective than others.
Nadia: Right, it means if a device only ever does one or two things repeatedly, an attacker has a much easier time building that profile against the noise.
Elias: So the implication is that for Matter developers, adding controlled variability to packet structures isn't just a suggestion; it’s becoming necessary defense against these passive traffic analysis attacks.
Priya: It sounds like we need to start designing protocols where structural randomness is a built-in feature rather than something you bolt on later.
Nadia: That’s what the researchers are pointing toward, making privacy a design requirement instead of an afterthought.
Elias: We can't just rely on the application encryption anymore; we have to consider how the communication itself is patterned when we think about security proofs and parameters.
Kristopher Alex Schlett, Béla Genge, Ioan Padurean, Savio Sciancalepore
Eindhoven University of Technology · Bitdefender
cs.CR
Submitted: 2026-02-02
Updated: 2026-11-03
Comments: Paper accepted to IEEE Consumer Communications & Networking Conference (CCNC) 2027
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
Importance score: 91/100
The gist: Matter is designed to enable interoperability across IoT products of various manufacturers, independently of underlying communication technologies.
Key concepts
- Matter Interoperability
- Matter is designed so devices from different manufacturers can work together without needing specific communication technologies. While this promotes compatibility, the consistent way devices communicate creates predictable patterns that an eavesdropper can analyze to understand how they interact.
- Encrypted Traffic Analysis
- Even when Matter traffic is encrypted at the application layer, passive attackers can infer device types and interactions by looking at characteristics like packet lengths and sequences. This analysis allows them to build a profile of the devices communicating without needing to decrypt the actual content.
- Packet Fingerprints
- The study established unique identifiers, or fingerprints, for different Matter device types (like sensors versus locks) based on specific packet structures, particularly IRA-1 lengths and sequences. These distinct patterns allow an AI system to accurately classify the type of device involved in a communication.
- Traffic Analysis Patterns
- The analysis identified numerous specific traffic patterns within encrypted data, including various RRA and RDA types. These consistent structural elements act as digital signatures that enable the system to reconstruct a timeline of commands and interactions even when the message content is hidden.
Terminology
Summary
Matter is designed to enable interoperability across IoT products of various manufacturers, independently of underlying communication technologies. p. 0 Things that Matter - Identifying Interactions and IoT Device
Types in Encrypted Matter Traffic
Kristopher Alex Schlett
Eindhoven University of Technology
Eindhoven, NL
Béla Genge, Ioan Padurean
Bitdefender
Targu Mures, RO
Savio Sciancalepore
Eindhoven University of Technology
Eindhoven, NL
Abstract
Matter is the most recent application-layer standard for the
Internet of Things (IoT). It is designed by the Connectivity
Standards Alliance (CSA) together with major commercial
players to enable interoperability across IoT products of various manufacturers, independently of the underlying communication technologies. However, to the best
of our knowledge, no existing research investigates the robustness of Matter-compliant communications to side-channel
attacks, particularly to encrypted traffic analysis performed
by passive actors possibly breaking the security or exploiting
the absence of network security of lower-layer technologies
(Wi-Fi, Thread). If successful, an adversary could use encrypted traffic analysis to infer which Matter interactions and
device types a user has, without even seeing the deployment.Where the paper says it
Where the paper says it
Research Question and Objectives
The main research question is what information a fully passive attacker can infer by analyzing encrypted Matter traffic. This overarching question is broken down into two sub-questions: RQ1 investigates how accurately a passive attacker can identify specific packets, transactions, and types of interactions from the analysis of encrypted Matter traffic. p. 2 We first
investigate the presence of recurrent patterns in the encrypted
Matter traffic. This is captured by RQ1:
RQ1: How accurately can a passive attacker identify specific packets, transactions, and types of interactions from
the analysis of encrypted Matter traffic? Moreover, we investigate to what extent such patterns
(RQ1) can be related to specific types of IoT devices in the
monitored environment. This is captured by RQ2:
RQ2: How accurately can a passive attacker infer the usage of specific device types from the analysis of encrypted
Matter traffic?Where the paper says it
Where the paper says it
Methodology and Data
The study utilizes extensive simulated and real-world data collected from heterogeneous environments to analyze the security of the Matter IoT standard to encrypted traffic analysis performed by a passive eavesdropper. The datasets include D0, which consists of encrypted Matter traffic (5M+ real-world Matter packets) collected over four weeks in E1, serving as a baseline to investigate the occurrence of patterns within communications involving specific device types. p. 4 The dataset D0 consists of encrypted Matter traffic (5M+ real-world Matter packets) collected over the
course of four weeks in E1, representing regular user-device
interactions and automations. It provides a comprehensive
collection of device interactions, thus serving as a baseline
to investigate the occurrence of patterns within communications involving specific device types. Format RQ
D0 Observe patterns in encrypted traffic E1 PCAP, HA Logs RQ1, RQ2
D1 Evaluate impact of delay and packet loss E3 PCAP RQ1, RQ2
D2 Infer interaction and label packets E2 PCAP, HA Logs RQ1, RQ2
D3 Test Device Classification E2 PCAP RQ2
Table 3: Datasets Description. Field Name Type Description
timestamp float (epoch seconds) Epoch time at which the packet was captured, expressed in microsecond precision
packet_len integer (bytes) Total packet size in bytes, including all protocol headers and payload data
src_id integer (identifier) Unique identifier of the device that generated/sent the packet
dst_id integer (identifier) Unique identifier of the device that is the intended recipient of the packet
ip_tos integer (0–255) IPv4 Type of Service (ToS) field indicating QoS and priority
ip_id integer IPv4 identification field used to identify packet fragments
ip_flags integer (bitmask) IPv4 flags controlling fragmentation (e.Where the paper says it
Where the paper says it
Key Findings on Interaction Identification
The analysis in dataset D0 discovered patterns that allow identifying various types of packets in the encrypted Matter traffic, including multi-attribute RRA, single-attribute RRA, empty RDA, non-empty RDA, End, WRA-1, WRA-2, IRA-1, IRA-2, SRA-2, and TRA. The methodology achieves an accuracy of 99. p. 8 72
IRA-1, RRA,
WRA-1 99. 66
Considering all patterns in Table 8, our methodology achieves
accuracy of 99. We report the details of the commands issued to
generate D2 in Table 6. To evaluate the extent to which we can infer interactions,
we investigate the following metrics: accuracy, recall, precision, and F1 score. Format RQ
D0 Observe patterns in encrypted traffic E1 PCAP, HA Logs RQ1, RQ2
D1 Evaluate impact of delay and packet loss E3 PCAP RQ1, RQ2
D2 Infer interaction and label packets E2 PCAP, HA Logs RQ1, RQ2
D3 Test Device Classification E2 PCAP RQ2
Table 3: Datasets Description. Field Name Type Description
timestamp float (epoch seconds) Epoch time at which the packet was captured, expressed in microsecond precision
packet_len integer (bytes) Total packet size in bytes, including all protocol headers and payload data
src_id integer (identifier) Unique identifier of the device that generated/sent the packet
dst_id integer (identifier) Unique identifier of the device that is the intended recipient of the packet
ip_tos integer (0–255) IPv4 Type of Service (ToS) field indicating QoS and priority
ip_id integer IPv4 identification field used to identify packet fragments
ip_flags integer (bitmask) IPv4 flags controlling fragmentation (e. To assess which interaction fingerprint best matches a given set of packets, we adopt nearestneighbor classification in set space [36], formalized through
the scoring metric in Eq. 1:
𝑆𝑥 = |𝑖𝑑 ∩ 𝑖𝑥 | + |𝑖𝑑 ∩ 𝑖𝑥 |
|𝑖𝑥 |
, (1)
where 𝑆𝑥 is the score for interaction type 𝑥, 𝑖𝑑 is the set
of observed packet lengths in the network traffic to the end
device 𝑑, and 𝑖𝑥 is the set of packet lengths in the fingerprint of
𝑥. As summarized in Table 5, we generate data considering two
distinct profiles of interactions per day, to represent various
scenarios. 𝐸𝑋𝑃1 represents realistic device usage patterns, i. For relatively low levels of packet loss, our
methodology is very robust. Indeed, it performs comparably
to the regular conditions reported in Table 9, still maintaining
over 94% and 93% in all performance metrics with an average
packet loss of 5% and 10%, respectively.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
Key Findings on Device Type Classification
Fingerprints of device types were established based on IRA-1 packet lengths and IRA-1/TRA packet sequences, revealing distinct fingerprints for lighting appliances, lock appliances, plug appliances, and sensors. When tested on dataset D3, the methodology achieves 100% for all four performance metrics in device type identification. p. 8 When focusing on the three most unusual
packet types, we notice a marginal decline in the investigated
metrics, except for precision (100%), indicating no false positives. Overall, in regular network conditions, our methodology
achieves strong performance in all labeling instances, indicating the feasibility of inferring specific device types from
encrypted Matter traffic.Where the paper says it
Discussion and Implications
The analysis shows that even though Matter is encrypted at the application layer, Read, Write, Invoke, and Report transactions remain identifiable through consistent packet lengths generated by Matter's design. This means that an eavesdropper can determine the device type and identify ongoing interactions despite encryption. p. 0 In this paper,
we analyze the robustness of the Matter IoT standard to encrypted traffic analysis performed by a passive eavesdropper. By analyzing various datasets collected from real-world and
simulated testbeds, we identify patterns in metadata of the
encrypted Matter traffic that allow inferring the interactions
occurring between end devices and controllers. Note that Matter is an applicationlayer protocol. Thus, proper investigation of its robustness
against encrypted traffic analysis must be performed considering full visibility of packet sizes, timings, and directions,
but no access to application payloads, in line with relevant
literature [33], [34].Where the paper says it
Where the paper says it
Conclusion
The results demonstrate that without robust countermeasures against traffic analysis, next-generation protocols like Matter risk undermining the privacy they aim to protect. The consistent patterns in encrypted traffic generated by Matter-enabled IoT devices allow for the inference of transactions and device types. p. 0 By analyzing various datasets collected from real-world and
simulated testbeds, we identify patterns in metadata of the
encrypted Matter traffic that allow inferring the interactions
occurring between end devices and controllers. Moreover,
we associate patterns in sequences of interactions to specific
types of IoT devices. Although Matter explicitly recommends enabling security
at the lowest layers of the protocol stack, MAC-layer security
in IoT protocols is often disabled, either by default by manufacturers or on-purpose by users to reduce overhead and energy consumption while increasing performance [8], [9], [10]. Even when enabled, network security relies on the secrecy
of user passwords: there is extensive research demonstrating
1
arXiv:2602.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 interaction classification in encrypted traffic allows an AI system to infer
specific Matter interactions (Read, Write, Invoke)
with over 99 percent accuracy by analyzing packet lengths and sequences found in D0. -
Device type identification capability enables an AI system to characterize the digital threat surface by inferring device types like
lighting, lock and plug appliances
with at least 88 percent accuracy from encrypted traffic patterns. -
The system can construct a detailed timeline of interactions where
commands can be explicitly identified despite encryption
by mapping observed packet lengths directly to command names based on the distinct fingerprints identified in Table 8.
Abstract
Matter is the latest interoperability standard for the consumer Internet of Things (IoT) market, developed by the Connectivity Standards Alliance (CSA). Matter's design pays particular attention to security and privacy: it provides validated secure session establishment protocols, and it uses robust security algorithms to encrypt communications between IoT devices and Matter controllers. However, little is known about the extent to which information can still be inferred from encrypted traffic. This paper investigates whether a passive adversary can identify Matter interactions and device types by analyzing encrypted traffic metadata. Using datasets collected from both real-world and simulated Matter deployments, we identify repeatable fingerprints based on packet lengths, directions, and packet sequences. Our results show that Matter interactions can be identified with over 95% accuracy even under packet loss and delay conditions. Furthermore, device types such as lighting, lock, and plug appliances can be inferred with accuracy exceeding 88%. These findings reveal a significant privacy risk affecting Matter-enabled deployments regardless of vendor. The CSA acknowledged our findings and indicated that mitigation mechanisms may be considered in future Matter versions.
Sources
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