Air-fi Rete locale Piergiorgio Venuti

Air-Fi: attacking computers that are disconnected and without network hardware is possible

Estimated reading time: 5 minutes

To keep secret information out of reach of attackers, organizations place it on devices that are not connected to any network. This is to avoid any possibility of communication with the Internet. These machines are called air-gapped . As safe as it may seem, infecting such a machine or network segment isn’t actually that difficult. Extracting the information obtained is much more difficult, but it was still possible with the Air-Fi technique .

To study an exploit of this scenario, all kinds of clever methods come into play, and Mordechai Guri, a researcher at Ben-Gurion University of the Negev (Israel), specializes in finding them. Dr. Guri is not the only one, of course, but in recent years, he has been involved in the discovery of a few dozen of these methods. Un new study describes how to extract data from an isolated computer, this time using Wi-Fi technology (hence the name Air-Fi ).

Air-fi Local network

How the Air-Fi method works

The beauty of Air-Fi is that it works even if the target computer has no Wi-Fi hardware. It relies on malware already installed on the device that can use the bus of DDR SDRAM memory to generate electromagnetic radiation at a frequency of 2.4 GHz . Malware can encode necessary data in variations of this radiation, and any device with a Wi-Fi receiver, including another compromised device, can collect and intercept the generated signals. This other device could be a regular smartphone or even a smart light bulb.

The Air-Fi method is particularly unpleasant from a cybersecurity point of view. It does not require administrator rights on the isolated computer; a normal user account can do the job. Also, using a virtual machine doesn’t provide any protection; VMs have access to memory modules.

Transmission range and speed

The researchers transmitted data without noteworthy distortion at a distance of up to 2-3 meters (in one case, up to 8 meters) and a speed of up to 100 bits per second , depending on the hardware of the infected computer and the type of receiver. Like most similar methods, it’s not very fast. Transferring a 20MB file would take 466 hours, for example. That said, the 1,300-byte “Jingle Bells” text could be transferred in 90 seconds. In this light, stealing a username and password with this technique seems entirely realistic.

Air-Fi RAM

How an attack could work

Infecting a air-gapped system with malware is not difficult. An attacker can easily do this by contaminating a USB drive, using social engineering or by tricking staff. Once done, the attacker would then have to infect a nearby WiFi-capable device to receive the leaked data. For this, the attacker can infect nearby desktops, laptops or even smartphones of personnel operating the target system with air-gapped .

To prevent this type of physical attack on the company, you may want to consider our service of physical test your company’s security !

After a successful infection, the malware steals data from the air-gapped system, leaking it into the air as Wi-Fi for the receiving device. As the researchers explained:

As part of the exfiltration phase, the attacker could collect data from compromised computers. The data can be documents, key records, credentials, encryption keys, etc. Once the data is collected, the malware starts the secret Air-Fi channel . It encodes the data and transmits it in the air (in the 2.4 GHz Wi-Fi band) using the electromagnetic emissions generated by the DDR SDRAM buses.

The following video shows a possible attack scenario.

The extraordinary absence of wi-fi hardware

As we have seen, the Air-Fi attack does not require specific Wi-Fi hardware to be installed on the target machines. How is it possible?

It is shown that the attack uses DDR SDRAM memory buses to generate electromagnetic emissions in the frequency band typical of the Wi-Fi protocol , ie 2.4 GHz Furthermore, it is also possible to encode data in binary code without specific privileges . Using a virtual machine doesn’t help, as they typically have access to hardware RAM anyway.

Communication between CPU and RAM modules takes place via a bus synchronized with the system clock . This generates electromagnetic radiation which will have a frequency related to the clock frequency. In the case of the DDR4 memory blocks it is around 2.4 GHz.

If the frequency of the modules is not the correct value, it is still possible to overclock or downclock the memory speed by adjusting it to the Wi-Fi frequency of 2.4 GHz.

In short, a machine that uses RAM blocks could still find a way to use them for data transmission. Of course, it all starts with a first compromise that installed malware on the machine.

How to defend yourself from Air-Fi

The use of Air-Fi involves electromagnetic emissions. It is possible to counter the strategy by using the following measures:

  • Do not allow Wi-Fi enabled devices to approach air-gapped systems for any reason
  • Monitor isolated systems for suspicious processes
  • Shielding the computer in a Faraday cage
  • Using SOCaaS to monitor networked machines
  • Control operations and visits to the company in order to eliminate the possibility of infection via USB stick

Like all similar methods, Air-Fi is too slow and difficult for common cybercriminals to use for everyday attacks. However, if your company is using air-gapped machines for data storage, it is certainly better to take cover, given the recent data hunger of cyber crime < / em>.

We recommend that you consider adopting a SOCaaS to prevent the use of malware, run regular procedures for verifying corporate security, both virtual ( Vulnerability Assessment & amp; Penetration Test ) and physical, as previously suggested, through our dedicated test service .

Contact us to find out how we can help you and how our services can secure your company data, we will be happy to answer any questions.

Useful links:

Share


RSS

More Articles…

Categories …

Tags

RSS Unknown Feed

RSS Full Disclosure

  • [NotCVE-2026-0001] Cloudflare Universal SSL CAA augmentation weakens RFC 8657 account binding — CVE-2026-14440 assigned 163 days after public no-CVE disclosure July 16, 2026
    Posted by NotCVE Advisories on Jul 15---------------------------------------------------------------------------- NotCVE Disclosure Update — NotCVE-2026-0001 / CVE-2026-14440 ---------------------------------------------------------------------------- [-] Summary: On 2026-01-19 the issue described below was published as NotCVE-2026-0001 after no CVE identifier was assigned for it. On 2026-07-01 — 163 days later — Cloudflare assigned CVE-2026-14440 to the same issue, now...
  • Subject: Advisory Submission: EZ Game Booster - Cleartext Storage of Sensitive Credentials (CWE-312) July 16, 2026
    Posted by AliReza on Jul 15# Exploit Title: EZ Game Booster v1.0.0 - Cleartext Credentials in user.config # Date: 2026-07-16 # Exploit Author: Alireza Chegini # Vendor Homepage: https://ezsystemrepairs.com # Software Link: https://ezsystemrepairs.com (Free version available) # Version: 1.0.0 (v2.0 exists but not tested - paid license required) # Tested on: Windows 10 / Windows […]
  • CVE-2026-56877 - Skillable SCORM userId authorisation bypass July 16, 2026
    Posted by Greg via Fulldisclosure on Jul 15Skillable&apos;s SCORM lab launch endpoint validates a launch token but enforces per-user allocation limits using a browser-supplied userId that is not bound to the validated token. An authenticated learner can modify this identifier to bypass configured limits, launch concurrent lab instances, and consume another learner&apos;s allocation. Skillable states […]
  • [REVIVE-SA-2026-003] Revive Adserver Vulnerabilities July 9, 2026
    Posted by Matteo Beccati on Jul 08======================================================================== Revive Adserver Security Advisory REVIVE-SA-2026-003 ------------------------------------------------------------------------ https://www.revive-adserver.com/security/revive-sa-2026-003 ------------------------------------------------------------------------ Date: 2026-06-25 Risk Level: Medium to High Applications affected: Revive Adserver Versions...
  • OPNsense XPATH Injection (CVE-2026-53582) July 7, 2026
    Posted by evan on Jul 06SUMMARY: a stored XPATH injection allows any user with just ca manager/certificate manager perms to leak any secret key/any value in config.xml, thus achieving privilege escalation and potentially remote code execution. this can also likely be chained via csrf and some clever hiding. see https://github.com/opnsense/core/security/advisories/GHSA-xww7-76m6-mh2r == VULN == the primary […]
  • SCHUTZWERK-SA-2025-001: Authentication Bypass for SafeLine SL6 and SL6+ July 7, 2026
    Posted by Jan Hüber via Fulldisclosure on Jul 06Authentication Bypass for SafeLine SL6 and SL6+ =============================================== The SafeLine SL6 and SL6+ devices integrated into elevator emergency intercom systems are vulnerable to an authentication bypass. This vulnerability allows attackers to bypass authentication requirements and access the device&apos;s configuration service via the Bluetooth Low Energy (BLE) interface. […]
  • Whistleblowersoftware.com: confidentiality and anonymity leakage to third parties July 2, 2026
    Posted by Red Nanaki via Fulldisclosure on Jul 02Whistleblowersoftware.com: confidentiality and anonymity leakage to third parties ## Summary The anonymous reporting flow on `whistleblowersoftware.com` encrypts report text end-to-end in the browser but does not extend the same guarantee to attachments and exposes the reporter&apos;s network identity and timing to a US-based third party. Together these […]
  • OpenBlow Multiple Deanonymization Vulnerabilities July 2, 2026
    Posted by Red Nanaki via Fulldisclosure on Jul 02OpenBlow Multiple Deanonymization Vulnerabilities Summary A production deployment was observed (HTTP archive of a full, real whistleblower submission) to route its anonymous reporting flow through Google. The intake CAPTCHA is Google reCAPTCHA, enforced as a mandatory, server-validated gate on report submission, and the UI additionally pulls a […]
  • Whistlelink: Site-access password exposed in web server access logs via GET query string July 2, 2026
    Posted by Red Nanaki via Fulldisclosure on Jul 02Whistlelink: Site-access password exposed in web server access logs via GET query string Severity: CRITICAL SUMMARY The Whistlelink reporting portal protects optionally-enabled, password-gated whistleblowing sites with a site-access password. When a visitor unlocks such a site, the client validates the password by issuing an HTTP GET request […]
  • APPLE-SA-06-29-2026-3 Safari 26.5.2 July 2, 2026
    Posted by Apple Product Security via Fulldisclosure on Jul 02APPLE-SA-06-29-2026-3 Safari 26.5.2 Safari 26.5.2 addresses the following issues. Information about the security content is also available at https://support.apple.com/en-us/127685. Apple maintains a Security Releases page at https://support.apple.com/100100 which lists recent software updates with security advisories. Web Extensions Available for: macOS Sonoma and macOS Sequoia Impact: A […]

Customers