Skip to main content
CybersecurityHacking

Telegram Exposes RCE Flaw in Android App

Smartphone and Android charging cable on a clean white surface with app icons and a blurred cityscape in the background.

"No one could be hacked just by receiving a message," Pavel Durov, Telegram's chief executive, wrote late last week — a claim that sits uneasily beside a fresh technical demonstration and vendor alarms about live exploitation.

NebuSec's demonstrations and claims

Security company NebuSec published a technical analysis and a demonstration this week showing two distinct ways a Telegram widget can run code outside the protections Telegram added to its Android app. The researchers ran their demonstration on a Telegram test server because, they note, only approved bots can send widgets on the live service at present.

In NebuSec's video, a bot's widget on Android 13.2.2 (7303) bypasses the app's new filter and opens the phone's calculator app. The write-up and demonstration do not claim what else the code might reach — for example, the app's messages or other phone data — nor do they say whether a user had to start the bot or tap the widget first. The researchers are withholding exploit details for now; their write-up lists no CVE identifier and does not say whether the bugs were reported to the app's developers.

What Telegram changed in Android 13.2.2 (7303)

After an account called V12 posted a video on X two days after bots gained the ability to send widgets, Telegram moved QuickJS — the small JavaScript engine that runs widget code — into the app's own process on Android and added a seccomp filter designed to limit system calls from that process. These changes appear in the published source for Telegram for Android 13.2.2 (7303).

The widget documentation states that widgets run in a sandbox with rules such as no network access; NebuSec notes those rules live in the engine's own code. Telegram previously told users that the problem in V12's first demonstration was fixed and that the fix had reached all users, and Pavel Durov said a user first had to start a third‑party service and that that vector was patched on the server side.

How NebuSec says the bypasses work

NebuSec says it found two separate bypass paths. One leverages a widget bug that the seccomp filter does not cover. The other chains a widget bug that the filter does cover with a kernel bug to escape protections. In general terms, the researchers explain that a single use‑after‑free or out‑of‑bounds write bug in QuickJS can let a script read and write the process's memory, after which the engine's sandbox rules no longer protect anything.

The researchers also point out architectural concerns: every widget runs inside a single process, so a hijacked widget from one bot would share memory with widgets from other bots. The new seccomp filter lists a few dozen system calls by name — including ptrace, bpf, and mount — and allows every other call through; it notably leaves ioctl unfiltered. The filter sits on top of Android's own limits for the engine process.

The published filter and audit mode

The Hacker News found the published source code for Telegram for Android 13.2.2 (7303) sets the new seccomp filter to an audit mode. In audit mode, the code will still block attempts to start new programs or processes but otherwise only records (does not block) calls listed in the filter and lets them run. A comment above the setting in the published source reads: "Audit until the logs from the devices say the engine never makes them."

The published source does not show whether release builds that users install use the same audit setting, and NebuSec's write-up lists no fixed version as of this week's publication.

Kaspersky's detection and bot‑policy note

Separately, security vendor Kaspersky said late last week it had detected suspected exploitation of a flaw in Telegram widgets when people visited some Telegram channels. In that activity, Kaspersky reported, a PowerShell script ran without the user noticing and downloaded and launched WebRAT, also known as Salat Stealer; the malware provides remote control and data‑theft capabilities aimed at crypto wallets and Telegram accounts.

Kaspersky did not identify the underlying flaw or the Telegram version, did not report victim counts, and did not connect the activity to either the V12 posting or NebuSec's demonstration. The vendor advised not running widgets until a fix is released. Meanwhile, a contributor to the telegram‑bot‑api project wrote in a GitHub thread that "Yes, currently, general bots aren't allowed to use widgets even in private chats," and the thread contains no later message lifting that restriction.

What this means for technologists, end users, and bot developers

  • Technologists and security teams: Review whether deployed builds of Telegram use the audit mode observed in the published source; audit mode records calls rather than blocking them, which affects the protection profile described in NebuSec's analysis.
  • End users: Exercise caution around widgets and third‑party bots; Kaspersky's detection shows at least one channel‑based activity that led to silent PowerShell execution and the deployment of WebRAT/Salat Stealer.
  • Bot developers and operators: Be aware that widgets run JavaScript via QuickJS inside the client process and that a single engine vulnerability can affect multiple bots because widgets share a process and memory space.

Telegram and the researchers offer competing reassurances: the company says the immediate server‑side vector was patched and that a user must initiate third‑party services, while NebuSec's demo and Kaspersky's telemetry show the attack surface remains worth close attention. The record in the published code raises a final, concrete question that only further disclosures will resolve: do released Telegram builds enforce the seccomp filter in blocking mode on user devices, or do they still run in audit mode that logs but permits the listed calls?

Original story at The Hacker News