Breaking
Quantum Computing

AI tool finds 99 new flaws in discontinued network devices

By Giulia Marchetti 3 min read
AI tool finds 99 new flaws in discontinued network devices - ai tool finds flaws
The tool analyzed open-source and compiled software to detect security issues in legacy network components.

A research team at Forescout has developed a tool to assess the ability of widely available AI models to detect security vulnerabilities on a limited budget. By examining both open-source and compiled software, whose source code is not accessible, the researchers evaluated an approach that could potentially be exploited by attackers. Their findings confirmed the existence of 99 vulnerabilities across 10 TCP/IP stacks, which are software components that enable devices to communicate over networks.

Comparing automated analysis with manual effort

The researchers discovered 66 previously unknown vulnerabilities, with 52% of these new findings rated as high or critical in terms of severity. Initially, the automated process generated 362 potential vulnerabilities, which were narrowed down to 312 candidates after removing duplicates. Following testing and review, 99 vulnerabilities were confirmed, including the rediscovery of 33 known vulnerabilities that were used for comparison purposes.

The cost of automated code inspection was approximately $935, while the total cost of the model, including estimated expenses for running software tests, ranged from around $1,200 to $2,300. In contrast, the researchers compared this to the roughly nine months of manual code analysis conducted during Project Memoria, a previous investigation that took place in 2020 and 2021. That project, which lasted 18 months, examined 14 networking stacks over multiple phases.

The automated analysis required around 10 hours, followed by approximately 80 hours of validation. The time and cost estimates are based on the application of the completed system to the target areas. Developing the system involved additional work, including the creation of scanning rules, preparation of environments for running software, and refinement of instructions for AI agents. The researchers utilized records from earlier investigations to develop 32 scanning rules, which covered issues such as buffer overflows and mishandling of network data.

Addressing vulnerabilities in legacy devices with no vendor support

The affected software is utilized in various products, including industrial equipment, medical devices, network equipment, and other connected devices. Siemens has identified affected building automation controllers in its APOGEE, Desigo, and TALON product lines, while other vendors continue to investigate. In one instance, the researchers analyzed firmware from a discontinued wireless access point made by a company that had ceased operations, discovering seven device-specific vulnerabilities, including unchangeable administrator credentials and a buffer overflow in the firmware upload process.

Organizations often face a difficult decision when dealing with legacy products that no longer receive official support: either continue using them or replace potentially mission-critical hardware. The Forescout researchers reported 54 newly discovered vulnerabilities to developers and maintainers, as they believed these issues still affected the latest versions. Several projects have since issued fixes, while investigations remain ongoing for three networking stacks whose names have been withheld during the disclosure process.

“Our research demonstrates that AI is changing the economics of vulnerability discovery and accelerating the pace at which security teams need to identify, prioritize, and mitigate cyber risk. This is especially impactful for critical infrastructure, where legacy and difficult-to-patch devices may remain in service for years,” said Daniel dos Santos, VP of Research at Forescout.

“Organizations still have legacy systems, operational requirements, and devices that can’t easily be patched or simply taken offline. Containing risk, rather than patching, must be the priority,” said Barry Mainz, CEO at Forescout.

Analysis of detected vulnerabilities by category

Testing confirmed the discovery of 33 previously known issues used for comparison. The remaining findings were split between memory-related errors and problems with network data handling.

Giulia Marchetti

Leave a Reply

Your email address will not be published. Required fields are marked *