Internet-Scale Measurement of React2Shell Exploitation Using an Active Network Telescope

arXiv:2603.12300 · cs.CR, cs.NI · Submitted 2026-03-12 · Read on arXiv

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: "Internet-Scale Measurement of React2Shell Exploitation Using an Active Network Telescope".

Elias: Here's the gist of the paper. The paper presents the first Internet-scale measurement study of React2Shell exploitation activity using traffic collected from an Active Network Telescope, which provides a quantitative characterization of scanning activity across geographic and autonomous system levels Vulnerability Context React2Shell is a critical remote code execution vulnerability (CVE-2025-55182) affecting React Server Components implementations, where insecure deserialization of component data exchanged through the React Flight Protocol enables remote code execution The flaw affects React Server Components implementations in versions 19.0.0, 19.1.0, 19.1.1, and 19.2.0, including the packages react-server-dom-webpack, react-server-domparcel, and react-server-dom-turbopack The maximum severity rating reflects the vulnerability’s potential to compromise application infrastructure and its suitability for large-scale Internet exploitation Exploitation activity has been observed shortly after public disclosure Measurement Methodology The study utilizes an Active Network Telescope (ANT) to gain insights into Internet security dynamics by monitoring unsolicited traffic directed toward unused IP address space A typical passive network telescope, though proven to be a powerful cyberspace observatory mechanism for inferring Internet-wide security dynamics, has a major limitation: it often does not receive application-layer data The authors augmented the conventional passive network telescope framework with active capabilities, responding to TCP SYN requests with appropriate SYN/ACK packets to capture the initial payload The proposed framework consists of four logical stages: capturing inbound traffic, reconstructing connection-level flows, decoding and normalizing payloads, and evaluating them using a protocol-aware detection signature Exploitation Detection Signature A conjunctive signature is adopted requiring the simultaneous presence of multiple exploitation intrinsic indicators within a decoded payload to classify it as React2Shell-related A payload is classified as React2Shell-related only if it satisfies all of the following conditions: first, the payload must contain an explicit reference to proto, reflecting an attempt to traverse inherited object properties during deserialization; second, the payload must include the string constructor: constructor, corresponding to chained access of the JavaScript constructor property; and third, the payload must contain a React Flight reference token matching the pattern +d+:.

Nadia: Only payloads exhibiting all three indicators concurrently are classified as React2Shell exploitation attempts Activity Characterization The analysis revealed rapid post-disclosure exploitation activity exhibiting patterns consistent with automated scanning campaigns, geographically distributed scanners, and concentrated backend infrastructure The first React2Shell exploit-bearing packet originated from Kerkrade, the Netherlands, at approximately 05:55:49 IST on 5 December 2025 and was recorded by VP1 Over the course of the month, it established 169,301 connections at VP1 and 43,108 connections at VP2 The campaign reached its peak on 19 December 2025, when VP1 recorded 88,197 React2Shell-related connections in a single day Infrastructure Asymmetry The analysis showed a clear separation between widely distributed scanning sources and a more concentrated set of backend server locations Approximately 37.1% of all observed traffic originated in the Netherlands and spanned multiple autonomous systems, indicating a non-centralized scanner infrastructure Backend server countries are identified by extracting IP addresses embedded within decoded payloads, which we interpret as server endpoints used for redirection or secondary-stage interactions during exploitation Collectively, approximately 90% of backend traffic was directed toward servers hosted within a small subset of countries, indicating a strong concentration of backend infrastructure supporting the observed exploitation activity Scanning Dynamics The study quantified the number of unique destination IPs contacted by each source IP at both vantage points A total of 735 distinct scanner IPs were observed at VP1, with a median number of destinations contacted per scanner being 32 The two time series exhibit strong temporal alignment, with a Pearson correlation coefficient of r = 0.87 (p < 0.01), indicating a statistically significant positive correlation This suggests that both vantage points captured consistent temporal dynamics of the campaign The proposed study reveals that exploitation activity emerged rapidly following public disclosure, with scanning behaviour exhibiting characteristics consistent with automated and coordinated probing Across the observation period, exploitation traffic showed clear temporal escalation after the vulnerability announcement, geographically distributed scanning infrastructure spanning multiple autonomous systems, and repeated payload structures indicative of scripted exploitation campaigns Despite the global distribution of scanners, the backend infrastructure referenced in exploit payloads was notably concentrated within a small set of server locations We further observe that exploit payloads overwhelmingly embed direct server IP addresses rather than domain names, indicating that the observed campaigns rely primarily on IP-based backend infrastructure Collectively, these observations highlight the rapid weaponization of critical application-layer vulnerabilities and the structural asymmetry between distributed scanning sources and comparatively centralized exploitation infrastructure Future work may extend this analysis by incorporating additional vantage points, correlating telescope observations with interactive honeypot deployments, and conducting longitudinal measurements to evaluate the persistence of React2Shell exploitation following patch adoption The results improve situational awareness of emerging web application vulnerabilities and underscore the importance of Internet-scale measurement for understanding the operational dynamics of exploitation campaigns targeting modern web frameworks Acknowledgments We acknowledge the Ministry of Electronics and Information Technology (MeitY), Government of India, and the CSIR Fourth Paradigm Institute (CSIR-4PI) for providing funding and scalable computational resources to support this research References

1: M.

Elias: F.

Nadia: S.

Paper summary: Nadia: We’ve seen how they measured the activity and what their methodology was. Now let’s talk about why this study matters in the bigger picture.

Elias: This paper, "Internet-Scale Measurement of React2Shell Exploitation Using an Active Network Telescope," is essentially providing the first internet scale measurement study of React2Shell exploitation activity >

Priya: It claims that this measurement helps characterize how fast the exploitation activity escalates after public disclosure, and it reveals patterns consistent with automated scanning campaigns and coordinated botnet activity >

Nadia: So, when we look at what they did, it wasn't just watching traffic; they used an Active Network Telescope to gain insights into Internet security dynamics by monitoring unsolicited traffic directed toward unused IP address space >

Elias: They augmented the conventional passive telescope framework with active capabilities by responding to TCP SYN requests with appropriate SYN/ACK packets to capture the initial payload >

Priya: The core of their finding is that they developed a conjunctive signature requiring the simultaneous presence of multiple exploitation intrinsic indicators within a decoded payload to classify it as React2Shell-related >

Nadia: Specifically, a payload is only classified as React2Shell-related if it contains an explicit reference to proto, which reflects an attempt to traverse inherited object properties during deserialization >

Elias: And it also has the string constructor, constructor, which corresponds to chained access of the JavaScript constructor property >

Priya: Plus, it must contain a React Flight reference token matching the pattern +d+:. Only payloads exhibiting all three indicators concurrently are classified as React2Shell exploitation attempts >

Nadia: It’s interesting how they tie together those specific indicators to define what an exploit looks like in this framework >

Elias: And they used synchronised measurements from two independent vantage points to demonstrate a strong temporal correlation in observed React2Shell activity, validating both their detection methodology and the exploit signature >

Priya: That temporal correlation is key because it confirms the absence of exploit-bearing traffic prior to public disclosure and captures the subsequent escalation and fluctuations in exploitation intensity >

Nadia: It helps fill those gaps regarding temporal evolution, geographic and autonomous system level distribution of attackers, and whether exploitation is driven by opportunistic scanning versus coordinated botnet activity >

Elias: So, this study is critical because it gives us empirical data on how this vulnerability actually behaves in the wild across the entire internet >

Conclusion: Nadia: Looking at the title of "Internet-Scale Measurement of React2Shell Exploitation Using an Active Network Telescope," what’s really going on with this work?

Elias: It boils down to using an active network telescope to measure exploitation activity for a specific vulnerability, React2Shell >

Priya: And the implication is that it moves us from limited anecdotal evidence to having a quantitative characterization of scanning activity across geographic and autonomous system levels >

Nadia: The paper shows that exploitation emerged rapidly following public disclosure with patterns consistent with automated and coordinated probing across multiple sources >

Elias: But the key structural asymmetry they found is between widely distributed scanning sources and a more concentrated set of backend server locations embedded in the exploit payloads >

Priya: It really highlights how these critical application layer vulnerabilities can be weaponized quickly, and it shows that while attackers might scan globally, their actual infrastructure for exploitation is often much more centralized >

Nadia: So, this study underscores the rapid weaponization of vulnerabilities and the structural difference between the distributed scanning sources and the concentrated exploitation infrastructure >

Elias: And it points toward a need to improve situational awareness regarding emerging web application vulnerabilities by measuring them at an internet scale >

Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, V. Anil Kumar

CSIR Fourth Paradigm Institute (CSIR-4PI) in Bengaluru, India

cs.CR, cs.NI

Submitted: 2026-03-12

Updated: 2026-11-09

Comments: arXiv admin note: This submission has been withdrawn because it does not meet arXiv's research content quality standards

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 90/100

The gist: The paper presents the first Internet-scale measurement study of React2Shell exploitation activity using traffic collected from an Active Network Telescope…

Key concepts

React2Shell
A critical remote code execution vulnerability (CVE-2025-55182) in React Server Components that allows attackers to execute arbitrary code by exploiting insecure deserialization of component data exchanged via the React Flight Protocol.
Active Network Telescope (ANT)
A monitoring system that actively probes the internet by responding to TCP SYN requests. This method captures application-layer data, unlike passive telescopes, allowing researchers to observe real exploitation attempts rather than just network flows.
Conjunctive Signature
A detection rule requiring three specific indicators in a payload simultaneously: an explicit reference to __proto__, the string 'constructor', and a React Flight reference token. Only payloads meeting all three criteria are classified as React2Shell-related exploitation attempts.
Infrastructure Asymmetry
The finding that while scanning sources are globally distributed, the backend infrastructure targeted in exploit payloads is highly concentrated in a small subset of countries. This reveals a structural imbalance between the attackers' probing methods and their actual exploitation targets.

Terminology

Summary

The paper presents the first Internet-scale measurement study of React2Shell exploitation activity using traffic collected from an Active Network Telescope, which provides a quantitative characterization of scanning activity across geographic and autonomous system levels

Vulnerability Context

React2Shell is a critical remote code execution vulnerability (CVE-2025-55182) affecting React Server Components implementations, where insecure deserialization of component data exchanged through the React Flight Protocol enables remote code execution The flaw affects React Server Components implementations in versions 19.0.0, 19.1.0, 19.1.1, and 19.2.0, including the packages react-server-dom-webpack, react-server-domparcel, and react-server-dom-turbopack The maximum severity rating reflects the vulnerability’s potential to compromise application infrastructure and its suitability for large-scale Internet exploitation Exploitation activity has been observed shortly after public disclosure

Measurement Methodology

The study utilizes an Active Network Telescope (ANT) to gain insights into Internet security dynamics by monitoring unsolicited traffic directed toward unused IP address space A typical passive network telescope, though proven to be a powerful cyberspace observatory mechanism for inferring Internet-wide security dynamics, has a major limitation: it often does not receive application-layer data The authors augmented the conventional passive network telescope framework with active capabilities, responding to TCP SYN requests with appropriate SYN/ACK packets to capture the initial payload The proposed framework consists of four logical stages: capturing inbound traffic, reconstructing connection-level flows, decoding and normalizing payloads, and evaluating them using a protocol-aware detection signature

Exploitation Detection Signature

A conjunctive signature is adopted requiring the simultaneous presence of multiple exploitation intrinsic indicators within a decoded payload to classify it as React2Shell-related A payload is classified as React2Shell-related only if it satisfies all of the following conditions: first, the payload must contain an explicit reference to proto, reflecting an attempt to traverse inherited object properties during deserialization; second, the payload must include the string constructor: constructor, corresponding to chained access of the JavaScript constructor property; and third, the payload must contain a React Flight reference token matching the pattern +d+:. Only payloads exhibiting all three indicators concurrently are classified as React2Shell exploitation attempts

Activity Characterization

The analysis revealed rapid post-disclosure exploitation activity exhibiting patterns consistent with automated scanning campaigns, geographically distributed scanners, and concentrated backend infrastructure The first React2Shell exploit-bearing packet originated from Kerkrade, the Netherlands, at approximately 05:55:49 IST on 5 December 2025 and was recorded by VP1 Over the course of the month, it established 169,301 connections at VP1 and 43,108 connections at VP2 The campaign reached its peak on 19 December 2025, when VP1 recorded 88,197 React2Shell-related connections in a single day

Infrastructure Asymmetry

The analysis showed a clear separation between widely distributed scanning sources and a more concentrated set of backend server locations Approximately 37.1% of all observed traffic originated in the Netherlands and spanned multiple autonomous systems, indicating a non-centralized scanner infrastructure Backend server countries are identified by extracting IP addresses embedded within decoded payloads, which we interpret as server endpoints used for redirection or secondary-stage interactions during exploitation Collectively, approximately 90% of backend traffic was directed toward servers hosted within a small subset of countries, indicating a strong concentration of backend infrastructure supporting the observed exploitation activity

Scanning Dynamics

The study quantified the number of unique destination IPs contacted by each source IP at both vantage points A total of 735 distinct scanner IPs were observed at VP1, with a median number of destinations contacted per scanner being 32 The two time series exhibit strong temporal alignment, with a Pearson correlation coefficient of r = 0.87 (p < 0.01), indicating a statistically significant positive correlation This suggests that both vantage points captured consistent temporal dynamics of the campaign

Conclusion

The proposed study reveals that exploitation activity emerged rapidly following public disclosure, with scanning behaviour exhibiting characteristics consistent with automated and coordinated probing Across the observation period, exploitation traffic showed clear temporal escalation after the vulnerability announcement, geographically distributed scanning infrastructure spanning multiple autonomous systems, and repeated payload structures indicative of scripted exploitation campaigns Despite the global distribution of scanners, the backend infrastructure referenced in exploit payloads was notably concentrated within a small set of server locations We further observe that exploit payloads overwhelmingly embed direct server IP addresses rather than domain names, indicating that the observed campaigns rely primarily on IP-based backend infrastructure Collectively, these observations highlight the rapid weaponization of critical application-layer vulnerabilities and the structural asymmetry between distributed scanning sources and comparatively centralized exploitation infrastructure Future work may extend this analysis by incorporating additional vantage points, correlating telescope observations with interactive honeypot deployments, and conducting longitudinal measurements to evaluate the persistence of React2Shell exploitation following patch adoption The results improve situational awareness of emerging web application vulnerabilities and underscore the importance of Internet-scale measurement for understanding the operational dynamics of exploitation campaigns targeting modern web frameworks

Acknowledgments

We acknowledge the Ministry of Electronics and Information Technology (MeitY), Government of India, and the CSIR Fourth Paradigm Institute (CSIR-4PI) for providing funding and scalable computational resources to support this research References

[1] M. F. S. Lazuardy and D. Anggraini, “Modern front end web architectures with react. p. 0

Where the paper says it

Index Terms—React2Shell, React, Network Telescope, Scanning, Security, Remote Code Execution Vulnerability I. Introduction MODERN web application development has increasingly shifted toward component-based architectures that integrate server-side rendering to improve latency, scalability, and search engine optimization.

js and next. p. 0
Where the paper says it

Introduction MODERN web application development has increasingly shifted toward component-based architectures that integrate server-side rendering to improve latency, scalability, and search engine optimization. Frameworks such as Next.

js,” Research Journal of Advanced Engineering and Science, vol. p. 0
Where the paper says it

While effective for performance, such protocols also expand the *Corresponding Author Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, and V. Anil Kumar are with the Big Data Research and Supercomputing Division, CSIR Fourth Paradigm Institute (CSIR-4PI), Bengaluru, India.

7, no. 1, pp. 132–141, 2022

[2] Y. Chen, Z. Wang, D. Whalley, and L. Lu, “Remix: On-demand live randomization,” in Proceedings of the sixth ACM conference on data and application security and privacy, 2016, pp. p. 1

Where the paper says it

In this paper, we use primary data collected from an Active Network Telescope (ANT) to gain insights into Internet security dynamics, specifically focusing on React2Shell incidents. Passive network telescopes monitor unsolicited traffic directed toward unused Internet Protocol (IP) address space, enabling the passive observation of large-scale scanning and exploitation activity without interacting with live production systems [15], [16].

50–61

[3] D. Abramov and A. Clark, “React Server Components,” ACM Queue, vol. p. 0

Where the paper says it

Anil Kumar∗ , Member, IEEE Abstract—The increasing adoption of server-side component-based web frameworks has introduced new application-layer attack surfaces that remain insufficiently understood at Internet scale. On 3 December 2025, a critical remote code execution vulnerability (CVE-2025-55182) in React Server Components, referred to as React2Shell, was publicly disclosed and subsequently observed being exploited in the wild.

21, no. 2, pp. 30–47, 2023

[4] Vercel, “Next.js app router and react server components,” 2024, technical documentation

[5] Meta Platforms, Inc., “The react flight protocol,” 2023, react technical documentation

[6] A. Tseng, R. Weiner, C. Charrier, Z. Work, G. Stark, and A. Larsen. (2025, Dec. p. 0

Where the paper says it

Anil Kumar∗ , Member, IEEE Abstract—The increasing adoption of server-side component-based web frameworks has introduced new application-layer attack surfaces that remain insufficiently understood at Internet scale. On 3 December 2025, a critical remote code execution vulnerability (CVE-2025-55182) in React Server Components, referred to as React2Shell, was publicly disclosed and subsequently observed being exploited in the wild.

) Multiple threat actors exploit react2shell (cve-2025-55182). p. 2
Where the paper says it

The framework reconstructs connection-level payloads from packet traces, applies deterministic multi-layer decoding, and isolates exploitation attempts using protocol-aware signatures grounded in the mechanics of CVE-2025-55182. • Using synchronised measurements from two independent vantage points, we demonstrate a strong temporal correlation in observed React2Shell activity, validating both our detection methodology and the exploit signature.

Google Threat Intelligence Group

[7] L. Davidson, “React2shell: Critical rce in react server components (cve-2025-55182),” https://react2shell. p. 0

Where the paper says it

Anil Kumar∗ , Member, IEEE Abstract—The increasing adoption of server-side component-based web frameworks has introduced new application-layer attack surfaces that remain insufficiently understood at Internet scale. On 3 December 2025, a critical remote code execution vulnerability (CVE-2025-55182) in React Server Components, referred to as React2Shell, was publicly disclosed and subsequently observed being exploited in the wild.

com, 2025, initial vulnerability disclosure report, November 29, 2025

[8] React Team, “Critical security vulnerability in react server components,” https://react.dev/blog/2025/12/03/critical-security-vulnerability-in-react-server-components, Dec. p. 0

Where the paper says it

Anil Kumar∗ , Member, IEEE Abstract—The increasing adoption of server-side component-based web frameworks has introduced new application-layer attack surfaces that remain insufficiently understood at Internet scale. On 3 December 2025, a critical remote code execution vulnerability (CVE-2025-55182) in React Server Components, referred to as React2Shell, was publicly disclosed and subsequently observed being exploited in the wild.

2025, security advisory for CVE-2025-55182 (React2Shell)

[9] Wiz Research, “Critical vulnerability in react (cve-2025-55182) – technical analysis,” https://www.wiz.io/blog/critical-vulnerability-in-react-cve-2025-55182, 2025, technical analysis, December 2025

[10] Cloudflare, “Mitigation actions related to react2shell exploitation attempts,” https://blog.cloudflare.com/react2shell-rsc-vulnerabilities-exploitation-threat-brief/, 2025, operational mitigation update, December 2025

[11] Cybersecurity and Infrastructure Security Agency (CISA), “Cisa adds cve-2025-55182 to known exploited vulnerabilities catalog,” https://www.cisa.gov/known-exploited-vulnerabilities-catalog, 2025, added December 8, 2025

[12] Z. Chen, C. Lever, and V. Paxson, “An empirical study of the log4shell vulnerability,” in Proceedings of the ACM Internet Measurement Conference, 2022

[13] M. Jonker and S. Kumar, “Internet-wide exploitation of log4shell,” IEEE Security & Privacy, vol. p. 0

Where the paper says it

Prior work has shown that high-severity framework vulnerabilities are rapidly weaponized following public disclosure. Measurement studies of Log4Shell demonstrated Internet-wide exploitation within hours, driven primarily by automated scanners and botnets [12], [13].

p. 0
Where the paper says it

Prior work has shown that high-severity framework vulnerabilities are rapidly weaponized following public disclosure. Measurement studies of Log4Shell demonstrated Internet-wide exploitation within hours, driven primarily by automated scanners and botnets [12], [13].

20, no. 4, pp. 14–23, 2022

[14] S. Kumar and A. p. 0

Where the paper says it

IEEE ACCESS 1 Internet-Scale Measurement of React2Shell Exploitation Using an Active Network Telescope Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, V. Anil Kumar∗ , Member, IEEE Abstract—The increasing adoption of server-side component-based web frameworks has introduced new application-layer attack surfaces that remain insufficiently understood at Internet scale.

Dainotti, “Characterizing exploitation of spring4shell at internet scale,” IEEE Security & Privacy, vol. p. 2
Where the paper says it

examined Internet-wide exploitation of Spring4Shell, revealing persistent scanning behaviour and delayed, yet sustained, exploitation attempts across multiple autonomous systems [14]. Earlier work on Shellshock also demonstrated how simple application-layer injection flaws can lead to large-scale compromise when deployed in widely used frameworks [26].

21, no. 3, pp. 52–61, 2023

[15] D. Moore, C. Shannon, and J. Brown, “Network telescopes: Observing the internet from the dark side,” Communications of the ACM, vol. p. 2

Where the paper says it

Moore et al. introduced network telescopes as a mechanism for observing Internet background radiation and large-scale attack activity [15].

49, no. 8, pp. 62–68, 2006

[16] A. Dainotti and A. King, “A survey of darknet and network telescope measurement,” IEEE Communications Surveys & Tutorials, vol. p. 0

Where the paper says it

IEEE ACCESS 1 Internet-Scale Measurement of React2Shell Exploitation Using an Active Network Telescope Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, V. Anil Kumar∗ , Member, IEEE Abstract—The increasing adoption of server-side component-based web frameworks has introduced new application-layer attack surfaces that remain insufficiently understood at Internet scale.

18, no. 1, pp. 137–157, 2016

[17] M. Allman and V. Paxson, “Ethical considerations in network measurement,” ACM SIGCOMM Computer Communication Review, vol. p. 2

Where the paper says it

Section II provides background on React Server Components, the React Flight protocol, and prior Internet-scale measurement studies of exploitation activity. Section III describes the ANT infrastructure, the data collection process, ethical considerations, and the detection pipeline used to identify React2Shell exploitation attempts in observed telescope traffic.

37, no. 4, pp. 65–72, 2007

[18] A. Dainotti and D. Moore, “Analyzing malware propagation using network telescopes,” IEEE Transactions on Network and Service Management, vol. p. 2

Where the paper says it

introduced network telescopes as a mechanism for observing Internet background radiation and large-scale attack activity [15]. Subsequent surveys and empirical studies have demonstrated their effectiveness for studying malware propagation, scanning campaigns, denial-of-service activity, and exploitation behaviour [16], [18].

9, no. 4, pp. 382–395, 2012

[19] H. B. Tanveer et al., “Unveiling ipv6 scanning dynamics: A longitudinal study using large scale proactive and passive ipv6 telescopes,” Proceedings of the ACM on Networking, vol. p. 1

Where the paper says it

In this paper, we use primary data collected from an Active Network Telescope (ANT) to gain insights into Internet security dynamics, specifically focusing on React2Shell incidents. Passive network telescopes monitor unsolicited traffic directed toward unused Internet Protocol (IP) address space, enabling the passive observation of large-scale scanning and exploitation activity without interacting with live production systems [15], [16].

3, no. CoNEXT3, pp. 1–24, 2025

[20] R. Stewart, “Stream control transmission protocol,” Internet Engineering Task Force (IETF), RFC 4960, Sep. p. 1

Where the paper says it

A typical passive network telescope, though proven to be a powerful cyberspace observatory mechanism for inferring Internet-wide security dynamics, has a major limitation: it often does not receive application-layer data. This is particularly problematic in the case of network activities, where the underlying transport is based on connection-oriented protocols such as Transmission Control Protocol (TCP), Stream Control Transmission Protocol (SCTP), and Multipath Transmission Control Protocol (MPTCP)[20], [21], [22].

2007, updated by RFC 6096, RFC 6335, RFC 7053, RFC 9260

[21] J. Postel, “Transmission control protocol,” Internet Engineering Task Force (IETF), RFC 793, Sep. p. 1

Where the paper says it

A typical passive network telescope, though proven to be a powerful cyberspace observatory mechanism for inferring Internet-wide security dynamics, has a major limitation: it often does not receive application-layer data. This is particularly problematic in the case of network activities, where the underlying transport is based on connection-oriented protocols such as Transmission Control Protocol (TCP), Stream Control Transmission Protocol (SCTP), and Multipath Transmission Control Protocol (MPTCP)[20], [21], [22].

1981, obsoleted by RFC 9293

[22] A. C. Ford et al., “Tcp extensions for multipath operation with multiple addresses,” Internet Engineering Task Force (IETF), RFC 8684, Mar. p. 1

Where the paper says it

A typical passive network telescope, though proven to be a powerful cyberspace observatory mechanism for inferring Internet-wide security dynamics, has a major limitation: it often does not receive application-layer data. This is particularly problematic in the case of network activities, where the underlying transport is based on connection-oriented protocols such as Transmission Control Protocol (TCP), Stream Control Transmission Protocol (SCTP), and Multipath Transmission Control Protocol (MPTCP)[20], [21], [22].

2020, rFC 8684

[23] M. Allman et al., “Tcp congestion control,” Internet Engineering Task Force (IETF), RFC 5681, Sep. p. 1

Where the paper says it

A typical passive network telescope, though proven to be a powerful cyberspace observatory mechanism for inferring Internet-wide security dynamics, has a major limitation: it often does not receive application-layer data. This is particularly problematic in the case of network activities, where the underlying transport is based on connection-oriented protocols such as Transmission Control Protocol (TCP), Stream Control Transmission Protocol (SCTP), and Multipath Transmission Control Protocol (MPTCP)[20], [21], [22].

2009, updated by RFC 9438

[24] A. Singh et al. p. 0

Where the paper says it

IEEE ACCESS 1 Internet-Scale Measurement of React2Shell Exploitation Using an Active Network Telescope Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, V. Anil Kumar∗ , Member, IEEE Abstract—The increasing adoption of server-side component-based web frameworks has introduced new application-layer attack surfaces that remain insufficiently understood at Internet scale.

, “A longitudinal measurement study of log4shell exploitation from an active network telescope,” arXiv preprint arXiv:2601. p. 0
Where the paper says it

0, there is limited empirical understanding of how this vulnerability is exploited across the Internet. This paper presents the first Internet-scale measurement study of React2Shell exploitation activity using traffic collected from an Active Network Telescope.

04281, 2026

[25] M. Jonker and S. Kumar, “Internet-wide exploitation of log4shell,” IEEE Security & Privacy, vol. 20, no. 4, pp. 14–23, 2022

[26] M. Antonakakis and T. April, “An empirical analysis of the shellshock vulnerability,” in Proceedings of the ACM Internet Measurement Conference, 2014

[27] R. Hiesgen et al., “Spoki: Unveiling a new wave of scanners through a reactive network telescope,” in 31st USENIX Security Symposium (USENIX Security 22), 2022, pp. p. 0

Where the paper says it

To the best of our knowledge, this work provides the first quantitative characterisation of React2Shell-triggered scanning activity, including the number of distinct scanners, their geographic and autonomous system distribution, and the scale of backend infrastructure involved in exploitation attempts. Index Terms—React2Shell, React, Network Telescope, Scanning, Security, Remote Code Execution Vulnerability I.

431–448

[28] R. Pang and V. Yegneswaran, “Inferring application-layer attacks from darknet traffic,” ACM SIGCOMM Computer Communication Review, vol. p. 2

Where the paper says it

Subsequent surveys and empirical studies have demonstrated their effectiveness for studying malware propagation, scanning campaigns, denial-of-service activity, and exploitation behaviour [16], [18]. Prior work has also shown that application-layer attacks, including HTTP-based probing and exploitation attempts, can be inferred from darknet traffic [28].

36, no. 4, pp. 67–78, 2006

[29] MITRE, “CVE-2025-55182,” 2025, available at https://cve.mitre.org/known-exploited-vulnerabilities-catalog

[30] Qualys Threat Research, “React2shell: Analysis of a critical react server components vulnerability,” Qualys Security Research, 2025

[31] Wiz Research, “Critical rce in react server components,” Wiz Security Blog, 2025

[32] Palo Alto Networks Unit 42, “Exploitation of react server components vulnerability,” Palo Alto Networks, Tech. Rep., 2025

[33] SecurityWeek, “Botnets exploiting react2shell vulnerability,” SecurityWeek, 2025

[34] Tenable Research, “React2shell: Detection and exploitation in the wild,” Tenable, Tech. Rep., 2025

[35] A. Felt and M. p. 3

Where the paper says it

Security of Web Application Frameworks and Serialization Mechanisms Academic research has long examined security risks in web application frameworks, including injection flaws, insecure deserialization, and abstraction-layer vulnerabilities. Felt and Finifter analysed systemic weaknesses in popular web frameworks, highlighting how framework complexity can obscure security boundaries [35].

Finifter, “Security analysis of modern web application frameworks,” IEEE Security & Privacy, vol. p. 3
Where the paper says it

Security of Web Application Frameworks and Serialization Mechanisms Academic research has long examined security risks in web application frameworks, including injection flaws, insecure deserialization, and abstraction-layer vulnerabilities. Felt and Finifter analysed systemic weaknesses in popular web frameworks, highlighting how framework complexity can obscure security boundaries [35].

9, no. 2, pp. 38–45, 2011

[36] C. Staicu and M. Pradel, “A study of prototype pollution vulnerabilities in javascript,” in Proceedings of the IEEE Symposium on Security and Privacy, 2019

[37] M. Johns and S. Lekies, “On the security implications of serverside rendering,” in Proceedings of the ACM Conference on Computer and Communications Security, 2018

[38] “IP2Location: IP Geolocation Database,” https://www.ip2location. p. 3

Where the paper says it

The extracted fields are stored within a structured relational database designed for scalable querying and correlation. We use IP2Location [38], a lookup service, to geolocate IP addresses and enhance contextual awareness by providing geographic and network attributes.

com/, accessed: 2025-01-26

Aakash Singh received the B.Tech. degree in Electronics and Communications Engineering from Guru Gobind Singh Indraprastha University, India, in 2021, and the M. p. 0

Where the paper says it

While effective for performance, such protocols also expand the *Corresponding Author Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, and V. Anil Kumar are with the Big Data Research and Supercomputing Division, CSIR Fourth Paradigm Institute (CSIR-4PI), Bengaluru, India.

Tech. degree in Artificial Intelligence from Netaji Subhas University of Technology, India, in 2023 Kuldeep Singh Yadav (M’20) received the Ph. p. 0
Where the paper says it

While effective for performance, such protocols also expand the *Corresponding Author Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, and V. Anil Kumar are with the Big Data Research and Supercomputing Division, CSIR Fourth Paradigm Institute (CSIR-4PI), Bengaluru, India.

D. degree in Electronics and Communication Engineering from the National Institute of Technology (NIT) Silchar, India Md. p. 0
Where the paper says it

While effective for performance, such protocols also expand the *Corresponding Author Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, and V. Anil Kumar are with the Big Data Research and Supercomputing Division, CSIR Fourth Paradigm Institute (CSIR-4PI), Bengaluru, India.

Talib Hasan Ansari received the B. p. 0
Where the paper says it

IEEE ACCESS 1 Internet-Scale Measurement of React2Shell Exploitation Using an Active Network Telescope Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, V. Anil Kumar∗ , Member, IEEE Abstract—The increasing adoption of server-side component-based web frameworks has introduced new application-layer attack surfaces that remain insufficiently understood at Internet scale.

Tech. degree in Electronics and Communication Engineering in 2016 and the M. p. 0
Where the paper says it

js [1] and Remix [2] leverage React Server Components (RSC) to enable fine-grained server execution of application logic while minimizing client-side computation [3], [4]. These architectures introduce new communication protocols, such as the React Flight protocol, which serializes component trees and state across trust boundaries between clients and servers [5].

Tech. degree in Electronics and Communication Engineering in 2018 from Jamia Millia Islamia, New Delhi, India V. p. 0
Where the paper says it

js [1] and Remix [2] leverage React Server Components (RSC) to enable fine-grained server execution of application logic while minimizing client-side computation [3], [4]. These architectures introduce new communication protocols, such as the React Flight protocol, which serializes component trees and state across trust boundaries between clients and servers [5].

Anil Kumar (M’15) is a Scientist G at the CSIR Fourth Paradigm Institute (CSIR4PI), Bengaluru Aakash Singh is currently a Scientist C at the CSIR Fourth Paradigm Institute (CSIR-4PI), Bengaluru, India Kuldeep Singh Yadav is currently a Scientist C at the CSIR Fourth Paradigm Institute (CSIR-4PI), Bengaluru, Gov. p. 0
Where the paper says it

While effective for performance, such protocols also expand the *Corresponding Author Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, and V. Anil Kumar are with the Big Data Research and Supercomputing Division, CSIR Fourth Paradigm Institute (CSIR-4PI), Bengaluru, India.

of India Md. p. 0
Where the paper says it

While effective for performance, such protocols also expand the *Corresponding Author Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, and V. Anil Kumar are with the Big Data Research and Supercomputing Division, CSIR Fourth Paradigm Institute (CSIR-4PI), Bengaluru, India.

Talib Hasan Ansari is currently a Scientist C at the CSIR Fourth Paradigm Institute (CSIR-4PI), Bengaluru, India since May 2024 V. p. 0
Where the paper says it

While effective for performance, such protocols also expand the *Corresponding Author Aakash Singh, Kuldeep Singh Yadav, Md Talib Hasan Ansari, and V. Anil Kumar are with the Big Data Research and Supercomputing Division, CSIR Fourth Paradigm Institute (CSIR-4PI), Bengaluru, India.

Anil Kumar is a Scientist G at the CSIR Fourth Paradigm Institute (CSIR4PI), Bengaluru Aakash Singh's research interests include web application security, artificial intelligence, geospatial analytics, high-performance computing, and applied machine learning for large-scale scientific and security applications Kuldeep Singh Yadav's research interests include cyber security, computer vision, affective computing, deep learning, image and video forensics, object detection, and pattern recognition Md. Talib Hasan Ansari's research interests include cyber security, computer vision, affective computing, deep learning, image and video forensics, object detection V.

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

  1. Detection of React2Shell exploitation attempts can be improved by implementing a deterministic, protocol-aware detection methodology that identifies structural markers absent from benign application traffic. This signature requires a conjunctive requirement: the simultaneous presence of multiple exploitation-intrinsic indicators within a decoded payload.

  2. The improved system can accurately characterize the temporal evolution of exploitation activity by analyzing how exploitation activity develops after disclosure, leveraging the data collected from Active Network Telescopes to capture the subsequent escalation and fluctuations in exploitation intensity.

  3. The system will enhance understanding of attacker infrastructure by characterizing "the geographic distribution of React2Shell activity at both the scanner and backend levels, revealing a clear separation between widely distributed scanning sources and a more concentrated set of backend server locations."

  4. The improved AI system can quantify the scale of backend infrastructure involved by analyzing the number of unique server IP addresses and autonomous systems, and their cumulative growth over time, thereby characterizing the structural asymmetry between distributed scanning sources and comparatively centralized exploitation infrastructure.

  5. The system will provide a more precise understanding of attacker behavior by analyzing destination-side behaviour by analysing the TCP ports targeted by exploitation attempts, highlighting a constrained and systematic focus on web-facing services.

Abstract

The increasing adoption of server-side component-based web frameworks has introduced new application-layer attack surfaces that remain insufficiently understood at Internet scale. On 3 December 2025, a critical remote code execution vulnerability (CVE-2025-55182) in React Server Components, referred to as React2Shell, was publicly disclosed and subsequently observed being exploited in the wild. Despite its critical severity and a CVSS base score of 10.0, there is limited empirical understanding of how this vulnerability is exploited across the Internet. This paper presents the first Internet-scale measurement study of React2Shell exploitation activity using traffic collected from an Active Network Telescope. We developed a deterministic detection methodology that identifies exploitation attempts targeting endpoints implementing React Server components. It helped analyze exploitation traffic to characterize its temporal evolution, geographic and autonomous system-level distribution, and behavioral properties of the observed scanning activity. In addition, exploit payloads are examined to understand the attacker infrastructure and delivery mechanisms. The analysis reported rapid post-disclosure exploitation activity exhibiting patterns consistent with automated scanning campaigns, geographically distributed scanners, and concentrated backend infrastructure. To the best of our knowledge, this work provides the first quantitative characterization of React2Shell-triggered scanning activity, including the number of distinct scanners, their geographic and autonomous system distribution, and the scale of backend infrastructure involved in exploitation attempts.

Sources

Related papers