NTU Agents Uncover 84 5G Core Vulnerabilities
On July 31, 2026, researchers at Nanyang Technological University reported 84 previously undisclosed flaws in open-source 4G and 5G core-network software. IT Security News reported that developers confirmed 83 findings, 81 received CVE identifiers, and 23 remained without fixes as of August 11. The most severe reported issue could enable subscriber-session hijacking, while the broader flaw class can also enable denial-of-service attacks.
Researchers at Singapore's Nanyang Technological University (NTU) reported 84 previously undisclosed security vulnerabilities in open-source 4G and 5G cellular-core software. According to IT Security News, developers confirmed 83 of the findings, 81 were assigned CVE identifiers, and 23 had no fix as of the outlet's August 11 report.
The findings concern a flaw class the researchers call implicit trust errors, or iTrues. The Hacker News reports that successful exploitation could enable denial-of-service attacks and, in a severe case, allow an attacker to hijack a subscriber's data session, causing the network to deliver the subscriber's traffic to the attacker rather than its intended destination.
Affected network components and protocols
The research paper, titled "Understanding Implicit Trust Errors in Core Carrier Networks through Multi-Agent Flaw Discovery and Analysis," examined LTE and 5G implementations across the GPRS Tunnelling Protocol Control Plane (GTP-C) and Packet Forwarding Control Protocol (PFCP) signaling interfaces.
According to The Hacker News, the examined projects included:
- •LTE: Open5GS and OpenAirInterface
- •5G: Open5GS, free5GC, OpenAirInterface, SD-Core, and eUPF
The researchers wrote that the recurring root cause was "implicit trust between core network functions." The Hacker News reports that vulnerable components may fail to validate message format, message semantics, or resource availability before acting on signaling messages.
LLM-assisted multi-agent discovery
NTU's team built an LLM-assisted multi-agent system to summarize known flaws, classify them into detection patterns, and use those patterns to identify new iTrues in core-network implementations, according to The Hacker News. The result is a concrete security-testing use case for agent systems: converting prior vulnerability knowledge into repeatable code and protocol-review workflows.
The reported exposure is especially relevant as cellular cores move from physically isolated infrastructure toward cloud-native deployments. The researchers told The Hacker News that this transition can make formerly internal interfaces reachable over the internet and weaken assumptions that components operate only within a trusted network zone.
For teams running private 4G/5G testbeds or integrating open-source core components, the findings underscore a broader engineering pattern: protocol parsing alone is insufficient when distributed services implicitly trust peer messages. Robust validation of sender identity, message state, semantics, and resource requests is necessary at inter-function boundaries, particularly where GTP-C or PFCP endpoints can be exposed beyond a tightly controlled network.
Key Points
- 1NTU identified 84 new flaws across open-source LTE and 5G cores, with 81 findings assigned CVE identifiers.
- 2The reported iTrue class stems from implicit trust among core functions, enabling denial-of-service attacks and potentially session hijacking.
- 3LLM-assisted multi-agent analysis found protocol-security defects, illustrating a repeatable approach for reviewing distributed telecom control-plane software.
Scoring Rationale
The report identifies a substantial set of confirmed flaws in open-source cellular-core implementations and includes a session-hijacking scenario. It is highly relevant to telecom security and agent-assisted vulnerability research, although the affected software is more specialized than general-purpose AI infrastructure.
Sources
Public references used for this report.
Practice with real Telecom & ISP data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Telecom & ISP problems
