Security researchers have documented six previously unknown vulnerabilities in Apple AirDrop and Google and Samsung Quick Share, showing that nearby attackers could crash transfer services or manipulate protocol state without any interaction from the victim. The work was presented at the 20th USENIX WOOT Conference on Offensive Technologies in August 2026 and resurfaced in November after a press release from CISPA, the German research centre where the work was done.
The findings are important, but they are also easy to overstate. The demonstrated impact is disruption of service and protocol abuse, not theft of files. The researchers did not show that an attacker could silently copy a victim's photos or contacts using these 2026 flaws. A Windows Quick Share use-after-free bug could theoretically be a route to code execution, but the paper does not present a full working exploit.
The study is a reminder that proximity protocols deserve the same scrutiny as internet-facing services. AirDrop and Quick Share feel like conveniences built into phones and laptops. Underneath, they are complex stacks of wireless networking, discovery, authentication and file transfer code. Any bug in those layers can be reached by someone in the same physical area.
What the researchers actually found
The paper, titled "Protocol Prying: Systematic Vulnerability Research in the Apple AirDrop and Android Quick Share Proximity Transfer Protocols," was written by Arash Ale Ebrahim and Nils Ole Tippenhauer of CISPA Helmholtz Center for Information Security. It was published in the USENIX WOOT '26 proceedings and later placed on arXiv.
The researchers reverse-engineered the AirDrop protocol stack and found three vulnerabilities in Apple's implementation. They found two in Samsung's Quick Share implementation and one in Google Quick Share for Windows. All six are reachable before a file transfer is accepted, which is why the paper describes them as pre-authentication issues in broad terms.
The AirDrop flaws can be triggered by sending crafted network requests while a device is discoverable. In the most straightforward case, an unknown HTTP path causes the sharing service to crash. Repeating the request roughly once every two seconds keeps the service unavailable. During the attack, the researchers reported that none of six legitimate AirDrop transfers succeeded.
The Quick Share findings are more varied. On Samsung devices, the researchers could manipulate protocol-state transitions before the encryption handshake was complete. On Windows, they found a race condition involving a use-after-free bug in the endpoint channel code. The Windows flaw was confirmed as a crash; the paper is inconsistent about whether full code execution was achieved.
Why proximity protocols are a distinct attack surface
Most security discussion focuses on attacks that arrive over the internet. Proximity protocols are different. They require the attacker to be close enough to participate in a local wireless exchange, often within roughly ten to thirty metres or on the same network. That limits scale, but it does not make the bugs harmless.
AirDrop and Quick Share are designed to make discovery and transfer frictionless. A phone advertises that it is available, nearby devices exchange metadata, and a recipient accepts a transfer. Each of those steps must parse data from another device before the user has approved anything. That pre-acceptance parsing is a rich source of bugs.
The attack model is also unusual because the victim may not know that anything happened. A crashed sharing service might simply look as though AirDrop or Quick Share has stopped working. A protocol-state manipulation might leave no visible trace at all. The user does not need to click a link or approve a request.
At the same time, proximity is a real constraint. These are not worms that spread across the internet. An attacker must be nearby, equipped with a compatible wireless device and able to send crafted traffic during a discovery window. That makes the findings most relevant to crowded public places, shared offices, conferences and transport hubs.
The AirDrop stack and the fuzzing approach
The research did not rely on guesswork. The team reconstructed AirDrop's layered architecture, from the Apple Wireless Direct Link network layer through HTTP, binary property lists, compressed archives and CPIO file structures. Understanding each layer was necessary because a malformed message at one layer might never reach the vulnerable code at another.
To automate discovery, the researchers built a protocol-aware fuzzer called AIRFUZZ. It is described as roughly 12,300 lines of code and uses pre-compression mutation, meaning it changes data before it is compressed rather than mutating compressed bytes blindly. That detail matters because compressed blobs are unforgiving: random changes usually produce invalid data that gets discarded early.
The fuzzer also used Frida hooks to observe internal behaviour. That let the researchers see which inputs were accepted and which code paths were reached. The result was a systematic search rather than a set of lucky crashes.
The artifacts were deposited on Zenodo with a DOI, which allows other researchers to inspect the tooling. That openness strengthens the work. It also means the findings are not merely a vendor claim or a press-release summary; other experts can reproduce and extend the analysis.
The three AirDrop crashes
The first AirDrop vulnerability affects the path router in Apple's Sharing framework. When the service receives an HTTP request for an unknown path, a Swift fatal error is triggered. The researchers demonstrated that this crashes the sharing daemon with a single crafted request, without any user action.
The second issue lies in the parser for Apple Foundation XML property lists. By nesting dictionary elements roughly 180 to 200 levels deep, the researchers caused a stack guard-page fault. This is a classic resource-exhaustion pattern: the parser assumes a bounded structure, but the input supplies something far deeper. The paper suggests similar code may exist across Apple platforms, but the reproduction was demonstrated on macOS and iOS.
The third flaw affects HTTP/1.1 handling in Apple's Network framework. A negative chunk size or conflicting Content-Length headers can cause a null-pointer dereference. This issue appears after a request has been accepted, which makes the broad pre-authentication label slightly less tidy, but it still does not require the victim to send a file.
The combined effect is a denial of service against AirDrop and related continuity features. Apple's sharing daemon is involved in more than AirDrop: AirPlay, Handoff, Universal Clipboard, Continuity Camera and NameDrop all rely on parts of the same proximity framework. The researchers argue that disrupting one component can affect several conveniences at once.
What Quick Share flaws showed
Quick Share is Google's answer to AirDrop, but Samsung and Google have implemented parts of it differently. The researchers examined a Samsung Galaxy S23 Ultra running Android 16 with Quick Share 13.8.01.11, and also Google Quick Share for Windows version 1.0.2472.1.
On Samsung devices, the first issue involves the dispatcher for offline frames. The researchers found that an OfflineFrame could be sent after a ConnectionRequest but before the UKEY2 encryption handshake. That opens a window in which protocol state can be manipulated. They reported five out of five successful reproductions at the affected stages.
A second Samsung issue concerns device-to-device frames accepted after the handshake. Three of seven frame types are handled as raw protobuf messages: connection response, bandwidth upgrade and keep-alive. By injecting these frames, an attacker could force the state to "accepted," inject keep-alive messages and cause the target to echo an attacker-supplied IP address and port. The researchers did not show that this enabled file theft.
The Windows issue is the most serious in principle. A use-after-free race in the endpoint channel can be triggered by colliding endpoint identifiers during concurrent connect and disconnect operations. The use of the identifier "FZ" was central to the proof of concept. The researchers confirmed a crash using WinDbg and Page Heap.
It is the Windows bug that raises the possibility of remote code execution. The binary involved reportedly had Control Flow Guard disabled, and the race window was about eleven seconds. But the paper does not provide a complete exploit. The Hacker News reported that the researchers confirmed a crash and suspected code execution but did not build a weaponised chain.
What attackers could and could not do
Precision matters here. The demonstrated attacks can break AirDrop and disrupt Quick Share interactions. They can force protocol states and cause crashes. A nearby attacker might be able to stop a transfer, interrupt a session or make a device's sharing features temporarily unusable.
The attacks do not, on the evidence presented, let someone steal files from a locked or unlocked phone. No contact database exfiltration, photo theft or message access was demonstrated in this 2026 research. The paper explicitly distinguishes its findings from earlier work on AirDrop privacy.
That earlier work is real and important. The 2021 USENIX Security paper PrivateDrop showed that AirDrop's contact discovery could expose phone numbers and email addresses through hashed identifiers, and that users could be tracked via an account-specific TLS identifier. Stute and colleagues had also examined Apple Wireless Direct Link privacy issues. Those findings should not be folded into the new WOOT paper as if they were one result.
The phrase "over five billion devices" also needs context. It comes from adding roughly 2.2 billion Apple devices and about 3 billion Android devices. That is an installed-base estimate, not a count of devices proven vulnerable to these six bugs. Affected versions were specific, and the attack range is local.
Patch status and the public-record gap
The disclosure story is more complicated than a simple fixed-or-not headline. According to the paper's disclosure section, Apple acknowledged the three AirDrop issues and said fixes were in progress, but no public CVEs had been assigned at that time. Samsung transferred two findings to Google. Google acknowledged the Windows issue, paid a bug bounty and said a CVE was pending.
Later reporting in June 2026 said one Apple bug had been fixed with a private, unpublished CVE, while Google had landed a code fix with a CVE still pending. The two Samsung issues remained under investigation. Then, in November, CISPA's press release stated that all six vulnerabilities had since been fixed.
That claim should be treated as vendor-reported. An inspection of Apple's public security advisory for iOS and iPadOS 26.5.2 does not show a public entry for AirDrop, sharingd, Foundation plist parsing or Network.framework. No public CVE or advisory could be found for the Quick Share Windows use-after-free as of 20 November 2026.
This does not mean the fixes are absent. It means the public record is incomplete. Private CVEs and silent patches are common in coordinated disclosure, especially when a vendor wants to avoid drawing attention to a flaw. But for users and enterprise defenders, the lack of a public advisory makes verification difficult.
Why the five-billion-device figure needs context
Large numbers make security stories travel, but they can mislead. The five-billion figure is a sum of two enormous ecosystems. It tells readers that AirDrop and Quick Share are widely deployed. It does not tell them that every device is affected by every flaw.
The AirDrop findings applied to specific macOS and iOS versions tested by the researchers, including macOS 15.7.3, 15.7.4 and 26.3, and iOS 18 and 26.3. The Samsung tests used a particular Galaxy device and software build. The Windows test used a specific Quick Share version. Other versions may behave differently, especially after patches.
The local range also changes the risk model. A vulnerability that requires proximity is less useful to a remote attacker than a flaw in a cloud service. On the other hand, local attacks can be very practical in physical settings where strangers are close together. Conference halls, trains and airports are exactly the places people use AirDrop and Quick Share.
The right conclusion is not panic and not dismissal. The flaws widen the known attack surface of everyday devices, but they do not mean every phone is constantly exposed to remote takeover.
How to reduce exposure today
The simplest precaution is to turn off AirDrop or Quick Share when it is not needed. On Apple devices, users can set AirDrop to "No One" rather than "Everyone for 10 Minutes" or "Contacts Only." The "Everyone for 10 minutes" setting was the condition used in several of the demonstrated AirDrop attacks.
On Android and Windows, users should keep Quick Share updated and disable it when not transferring files. Restricting discovery to contacts, where the platform offers that option, also reduces the number of people who can send crafted requests. In a high-risk setting such as a security conference, leaving proximity sharing off entirely is reasonable.
Installing vendor updates remains important even when advisories are sparse. The CISPA statement says the issues have been fixed, and independent reporting says at least some patches have shipped. Users cannot verify every CVE themselves, but keeping operating systems and sharing apps current is still the best default.
Enterprises can go further. Device-management policies can disable proximity sharing for managed devices or restrict it to approved contexts. Security teams should not treat these flaws as a reason to ban all local sharing outright, but they should include proximity protocols in threat models.
What this means for protocol design
The study's deeper lesson is that convenience features need adversarial parsing from the start. Every pre-authentication input is a potential attack surface. If a service must parse a property list, an HTTP header or a protobuf frame before the user has approved a session, that parser must be designed for hostile data.
Resource limits matter. Deeply nested structures, unbounded frame types and complex state machines all create opportunities for crashes. The AirDrop plist issue is a good example: a parser that assumes reasonable nesting will eventually meet an unreasonable input.
State machines need equal care. The Samsung findings show what can happen when a message arrives out of the expected order. A protocol that treats a frame as valid before the encryption handshake is complete has effectively moved a trust boundary forward. Designers should enumerate every possible state transition and ask what happens when messages arrive early, late or twice.
Fuzzing is clearly useful. AIRFUZZ found issues that manual review alone might have missed. But fuzzing works best when it knows the protocol. The researchers' decision to mutate before compression was central to getting accepted inputs deep into the stack. That is a practical lesson for anyone testing complex file-transfer systems.
The broader lesson for users
The vulnerability story is ultimately about the hidden complexity of ordinary features. AirDrop and Quick Share look like simple buttons. They are actually negotiations between devices over radio links, identity checks, compression formats and file archives. Each layer has to trust the one below it, and each layer can be attacked.
The good news is that this research was conducted responsibly. The findings went through academic peer review at USENIX WOOT, artifacts were published, and vendors were notified. The bad news is that the public patch record is murky. Users are asked to trust that fixes have landed without always being able to see exactly which CVE corresponds to which bug.
For now, the practical advice is modest: keep systems updated, turn off proximity sharing when it is not needed, and avoid using "Everyone" discovery modes in crowded places. The research does not justify claims that AirDrop or Quick Share are secretly stealing data. It does justify a more careful look at the small wireless doors we leave open.
Nexuswild welcomes factual corrections. Email contact@nexuswild.com with evidence and the article URL.
