TRENDING
Rows of red climbing helmets, carabiners and harness straps laid out on the ground at a ropes course
October 21, 2026
How to Build a Safe AI Agent Harness in Python With Tool Validation, a File Sandbox, and Run Limits
A thin transparent polymer test specimen stretched between the metal grips of a tensile testing machine
October 21, 2026
Ai2’s FairStress Turns Fairness Testing Into a Pressure Test, but the Released Files Do Not Match the Report
A vintage Banda Model Ten spirit duplicator, a hand-cranked paper copying machine, shown with its instruction book on a white background
October 21, 2026
An Anonymous Comment Can Hand Duplicator a Payload That Fires at Migration Time, and Under a Third of Free Installs Run the Fix
Colorful horizontal layers of sedimentary rock exposed in a cliff face at Makhtesh Ramon in Israel, with two desert shrubs in the foreground
October 21, 2026
How to Build an LSM-Tree Key-Value Store in Python With a Write-Ahead Log, SSTables, Bloom Filters, and Compaction
A broadcast studio teleprompter with script lines shown mirrored on its lower monitor and reflected in the glass in front of a camera lens
October 21, 2026
WhatsApp’s Leaked Chatbot Notes Turn a Help Page Into Two Documents, and the Hidden One Told the Bot Not to Recommend a Regulator
21 Oct 2026
SXZ.io SXZ.io
  • Home
Search the Site
Popular Searches:
Technology Amazon AI
Recent Posts
An out-of-order sign standing in an airport sink basin, with a title, a friendly explanation and a Spanish translation, like the members of an error response
How to Return RFC 9457 Problem Details From a FastAPI API and Stop Error Responses From Leaking Secrets
October 21, 2026
Heavily corroded cogwheels and a rusted ring gear inside an abandoned sulphur factory beside the sea on White Island
Cloudflare’s tsrs Turns an Agent-Written Compiler Port Into a Differential-Testing Bet, and the 8x Headline Depends on the Baseline
October 21, 2026
A black public warning siren with a wide dish-shaped horn on a short pole against a clear blue sky
PHP Warns of Two Critical Flaws and Sets a 14:00 UTC Patch, but the Leak, the Score and the Exploitation Reports Do Not Agree
October 21, 2026
SXZ.io SXZ.io
  • Home

Categories

Articles 255 Posts
News 258 Posts
Learning Hub 228 Posts
Home/News/Targeted Ads Delivered Candiru Spyware to Al Jazeera, Citizen Lab Says, but Neither Report Describes the Exploit
News

Targeted Ads Delivered Candiru Spyware to Al Jazeera, Citizen Lab Says, but Neither Report Describes the Exploit

Citizen Lab says ads bought by Discover-Tech, a vendor Google and IAB Europe listed since 2020, infected Al Jazeera devices with Candiru spyware. Neither it nor Confiant describes the exploit, so...

October 15, 2026 11 Min Read
53

Citizen Lab reported on Thursday, October 15 that targeted online ads infected devices at Al Jazeera with spyware from Candiru, which now operates as Integrity Labs, and that at least 45 devices at the broadcaster were targeted in 2024 and 2025. The researchers call the technique AdFection and describe their report as the “first-ever public confirmation of zero-click mercenary spyware delivered via targeted ads.” The ads were bought by Discover-Tech, an Israeli company that presents itself as an advertising vendor and that, Citizen Lab says, works as a facilitator for several spyware firms. Google and IAB Europe, which had listed the company as a certified or approved vendor since 2020, removed or suspended it in early October after the researchers disclosed their findings.

Table Of Content

  • What Citizen Lab found
  • How big it was
  • What the evidence shows and what stays open
  • How the ads got past ad verification
  • Who vouched for the vendor
  • What defenders can do
  • What to watch
  • Sources and method

The reports leave a gap that matters for defenders. Confiant, an ad-security firm that published a companion analysis the same day, says it has “not captured a full exploit chain in the wild,” and Citizen Lab’s report neither describes the exploit that did the infecting nor names a vulnerability behind the Al Jazeera infections. So the reports give defenders no specific vulnerability to patch against. What the researchers do have is network logs from Al Jazeera, a malware sample from one compromised device and a detailed view of the vendor’s infrastructure, which together are enough for Citizen Lab to attribute the infections to Candiru with “high confidence.”

What Citizen Lab found

Discover-Tech runs an ad-buying platform called MyDSP and also buys ads through Beeswax, which Comcast’s FreeWheel owns, and through Eskimi. Many of its ads carry a script tag that calls back to what the company names a NOAS server. The ad reports the viewer’s cookie ID or mobile advertising ID, and the server decides, apparently from that identifier and the IP address, whether the viewer is eligible to receive code. Most viewers are not and get an empty response. Citizen Lab says that design lets Discover-Tech avoid “directly sharing malicious code, or even pointers to malicious code, with the companies who serve and display the ads.”

A second Discover-Tech platform, Controlium, works as the control panel. According to Citizen Lab, customers choose whom to profile or infect by mobile advertising ID (iOS or Android), browser ID or IP address. Discover-Tech published the front-end source code on the login page, which let the researchers rebuild the interface and read its screens.

The Al Jazeera case rests on network logs that the broadcaster shared with Citizen Lab, and Al Jazeera agreed to be named in the report. Citizen Lab says infected devices first contacted Discover-Tech domains, which it reads as an ad impression, then connected to a Candiru validator server that checks the device before it is profiled and infected, and finally contacted a Candiru command-and-control server. Al Jazeera also isolated a Windows DLL from one compromised device. The sample hard-codes two command servers that match Citizen Lab’s fingerprint, persists through a COM hijack, hides strings with a 48-byte repeating XOR key and checks connectivity against google.com and microsoft.com, the same traits as the Candiru chain the lab captured in 2021. Citizen Lab calls the sample a newer version of that chain. WIRED reports that Al Jazeera did not respond to its request for comment.

Discover-Tech also hands each target’s details to the attacker in encrypted form. In Candiru infection links from 2024 and 2025, a short query parameter decrypted every time with a key published on an “Ad URL OpSec Decoder” page on a Discover-Tech-linked website. Citizen Lab reads this as a secure handoff: the attacker can compare the visitor’s IP address with the one Discover-Tech encrypted, so a researcher cannot swap the link onto another device to capture an exploit. In one link posted to VirusTotal the value decrypted to an IP address registered to the Saudi Ministry of Foreign Affairs, which the report takes as an indication of a potential target.

How big it was

Citizen Lab found two world-readable Apache Druid databases on a Discover-Tech domain that exposed MyDSP bidding records until Discover-Tech appeared to shut them down in August. From March to June 2026, MyDSP bid on 3,770,286,576 impressions and appeared to win 441,954,291, or 11.7 percent. Yandex carried 67.4 percent of the bids, Google Ad Manager 24.74 percent and Vungle 7.24 percent. The ads ran in more than 70,000 mobile apps and more than 120,000 websites. The records list three advertisers, most of them Discover-Tech or entities linked to it, and the conversion fields are always zero, which the report calls highly unusual for an ad vendor while allowing that the records might not track conversions. The records put spend at about $109,091 if, as the report assumes, cost is stored in micros of US dollars. By our arithmetic that is roughly 25 cents for every thousand impressions won.

Citizen Lab says the ads also “covertly directed millions of devices to contact Candiru’s servers, likely for large-scale tracking and reconnaissance.” It estimates that between hundreds of thousands and millions of users were sent to Candiru-linked domains during 2025, and it counted 3,637,213 “eligible users” on a Discover-Tech server from January to June 2026, while cautioning that other Discover-Tech customers may use the same feature. The Register notes that those connections do not necessarily indicate spyware infections.

The same data show Discover-Tech ads reaching devices on the networks of defense ministries, armed forces, national cyber agencies, NATO, the United Nations and EU institutions, among others. The report says an impression “does not in and of itself confirm targeting or infection,” but that “at minimum” these impressions “would have resulted in profiling by Discover-Tech.” The Register summarizes the finding as researchers having “found no evidence that those organizations were infected.” The report’s wording is more open: it says the data “warrants urgent investigation to determine whether any users on these networks were infected with spyware,” and it notes that its view covers only MyDSP data, not Beeswax or Eskimi.

What the evidence shows and what stays open

Claim What the reports show What stays open
Al Jazeera devices were infected through ads Logs show Discover-Tech domains, then a Candiru validator, then a Candiru command server. A Windows DLL from a compromised device matches the 2021 Candiru chain. Neither report describes the ad-delivered exploit. For the Candiru-linked URLs it traced, Citizen Lab was “unable to obtain any of the JavaScript code.”
Candiru ran the infections Server fingerprints (305 validator domain names on 319 IP addresses) and the DLL, rated “high confidence.” Candiru did not respond. Citizen Lab calls the status of its reported 2025 acquisition by Integrity Partners unclear.
Discover-Tech exists to link ad vendors and spyware firms The Controlium panel, NOAS gating, the encrypted target handoff, and Controlium sign-ins from users Citizen Lab links to Candiru, Bindecy and Intelos. Discover-Tech did not respond to Citizen Lab, WIRED or The Register. The report says the ad platforms it used were “not necessarily aware” of misuse.
Millions of devices were profiled Ads covertly sent millions of devices to Candiru servers; 3,637,213 eligible users from January to June 2026. A connection is not an infection, and the count may include other Discover-Tech customers.
Sensitive networks were exposed Impressions on networks of defense ministries, NATO, the UN and EU bodies in the MyDSP data. An impression does not confirm targeting or infection, and the Beeswax and Eskimi data were not available.
There is a vulnerability to patch Neither report names one for the Al Jazeera infections. The CVEs in Citizen Lab’s text belong to Candiru cases in 2021 and 2022. The reports do not say which browser or operating-system flaw ran after the ad loaded.

How the ads got past ad verification

Eskimi, one of the platforms Discover-Tech bought ads through, told Citizen Lab that the advertiser’s tags “were built to evade automated scanning. They loaded different code depending on who was viewing the ad, including the device, IP address and location, and campaigns were often restricted to very narrow IP target lists. This is how they passed third-party verification.” Eskimi says Discover-Tech had been a self-service advertiser on its platform since 2022, spent only a few hundred dollars, and had its last campaign stopped on August 6, 2026.

Confiant traced the same behavior from the other end. Its scanner flagged a Discover-Tech ad on April 12, 2026, and its analysis covers April 1 to September 10. The ad’s script hands off to an obfuscated JavaScript SDK of about 90 KB that fingerprints the device, watches mouse, touch and keyboard behavior, sets up an encrypted channel to a command server and polls it once a second. The operator can push code that the SDK runs with a new Function call, in the ad frame. Confiant calls that “an effective mechanism for the surgical delivery of high-impact payloads such as browser exploits,” then states the limit: “Even though we have not captured a full exploit chain in the wild, the surrounding staging demonstrates an advanced capabilities package built to evaluate targets before delivering high-impact payloads.” A second payload carries a Google copyright comment that Confiant calls a decoy, a trick ESET documented in Candiru JavaScript in 2021.

Confiant also describes a separate harvest. In ad sessions across more than 1,000 publisher domains, Discover-Tech creatives sent the page’s first-party cookies back to the operator, and in many cases they included session IDs of logged-in users and, sometimes, emails and full names inside JWT tokens. Confiant says it saw this even where Discover-Tech lacked GDPR consent. By its description the siphon works only when the ad runs in the publisher’s own origin, either in a friendly iframe or written into the page, and only for cookies not marked HttpOnly. In a cross-origin SafeFrame it returns little or nothing. That points publishers to two controls, our reading rather than Confiant’s advice: mark session cookies HttpOnly and prefer cross-origin ad frames.

Confiant’s telemetry also shows the vendor’s ads concentrated unevenly. It counted Discover-Tech serving 1.9 percent of all ads in Palestine during the period, peaking at 9.9 percent on April 7, including 1 in 36 Android ads there, and 1 in 54 iOS ad impressions in the Netherlands. Confiant says the data broadly supports deliberate targeting by device type and operating system.

Who vouched for the vendor

Citizen Lab’s Table 9 shows Discover-Tech in IAB Europe’s Transparency and Consent Framework Global Vendor List as vendor 778 from July 2, 2020, and in several Google lists that also start in 2020, among them roughly 2,400 certified external vendors for Ad Manager. Google says certified vendors “undergo a business review,” and Citizen Lab notes the nature of that review is not specified. Its verdict on the regime is blunt: the industry “appears to perform no meaningful due diligence on participating vendors, instead improperly relying on self-attestation where vendors check a box on an online form.”

We checked two of the lists at about 19:10 UTC on October 15. The Global Vendor List (version 181, last updated 16:00 UTC today) still contains vendor 778, Discover-Tech ltd, but with a deletedDate of 2026-10-08T13:57:02Z. The framework’s specification says a vendor with that field “is considered deleted after this date/time and MUST NOT be established to users.” WIRED reports that IAB Europe says it suspended the company after an audit of its policies, and quotes a spokesperson saying “further confirmation of the findings regarding spyware could also justify a definitive expulsion from the GVL.” So the status is a suspension in IAB Europe’s words and a deletion flag in the file. Google’s ads personalization providers file no longer has provider 3036 or any row that names Discover-Tech, mtrtb or mydsp.

Other platforms responded to Citizen Lab over the past week. Google says it “terminated Discover-Tech’s access to its ad exchange platforms, blocking them from bidding or serving ads across our ecosystem” and added Safe Browsing warnings. Beeswax says it removed Discover-Tech from its buying platform after third-party validation, Vungle (Liftoff) paused all of its access during a review, and Eskimi blocked its accounts and domains. Adtimizer and Volasim denied involvement or knowledge. Xandr (Microsoft Advertising), Sunteng, IAB Europe and Affle did not respond to Citizen Lab within its deadline.

What defenders can do

Citizen Lab’s recommendations reduce exposure to ad code rather than fix a bug. For organizations it advises blocking ads and tracking at the gateway and on work devices, opting company-issued devices out of ad targeting (it mentions MDM as one route), preferring browsers with built-in anti-tracking and Global Privacy Control support, and watching traffic tied to ad impressions for malicious requests. People at elevated risk should seek expert help and can reset their Mobile Advertising ID. The report warns that blocking advertising identifiers stops Discover-Tech’s current methods, but that “the AdFection attack surface may be larger.” Our September post on the Pentagon’s ad-tracker ban shows a limit of the identifier advice: lawmakers asked why disabling advertising identifiers had not stopped data brokers from exposing troop locations.

The advice to block ads has a catch today. At 19:13 UTC we searched the text of EasyList, EasyPrivacy, uBlock Origin’s filters, privacy and badware lists, and Peter Lowe’s list for the domains named in the reports, and none contained a rule for them. AdGuard’s DNS filter listed two, tracker.mtrtb[.]com and myweb3id[.]io, but nothing for the ppv2, noas or infinoas hosts that Eskimi names as loaders, and the one StevenBlack hosts match, controliumbt[.]com, is a different domain from the names in the reports. We searched the lists by name and did not test them in a browser, so a generic pattern rule could still catch some requests.

DNS lookups at 19:18 UTC show which names still answer. We queried 1.1.1.1 and 8.8.8.8, looked up registry records, and connected to none of these hosts.

Name Role according to the reports Status at 19:18 UTC
mtrtb[.]com (ppv2, noas, infinoas hosts) Discover-Tech ad server and loader hosts Resolves
events.myweb3id[.]io Host named by Eskimi among those its blocked ad tags called Resolves
myipinfo[.]io Discover-Tech-linked advertiser site that hosts the Ad URL OpSec Decoder Resolves
pagead.trustlayersdk[.]com Command endpoint of the SDK Confiant analyzed Resolves
dtech.adtimizer[.]co Loader host named by Eskimi NXDOMAIN
keyadapter[.]com, sharethere[.]net Candiru validator servers NXDOMAIN
fearevolve[.]com, jobmarcher[.]com Command servers hard-coded in the Al Jazeera DLL NXDOMAIN
tracookiepixel[.]xyz, cookie-sync[.]xyz Cookie-matching domains Confiant says were suspended on October 1 NXDOMAIN; registry status “server hold,” last changed October 1 at 18:21 UTC

Resolving does not mean serving malicious code, and the reports hold fuller indicator lists. The point is narrower: the names at the center of the ad-serving chain were still live hours after the disclosure, so a DNS or gateway block is still a control you can apply.

What to watch

Citizen Lab says a second report will cover DARK CARTOGRAPHER, a separate actor using Discover-Tech infrastructure to link advertising IDs to phone numbers by exploiting vulnerabilities across mobile carriers, which it says has run since at least 2019. It will publish once disclosure to providers is complete. The lab also asks regulators to investigate Discover-Tech and Candiru and to require third-party audits and know-your-customer checks for vendors in “certified” registries. Candiru was added to the US Commerce Department’s Entity List on November 3, 2021, and the report says Israeli media reported in April 2025 that Integrity Partners had acquired its operations. For how little audit sits behind spyware vendors’ promises, see our October 1 post, where Paragon’s CEO said the company cannot see who its customers target.

Sources and method

We read the full text of Citizen Lab’s report, including its correspondence section, and Confiant’s analysis, plus the coverage in The Register and WIRED. Every figure about bids, spend and impressions comes from Citizen Lab’s analysis of the exposed Druid data, and the roughly 25 cents per thousand impressions is our division of two of its numbers. The registry checks used IAB Europe’s published vendor list and Google’s published providers file. The filter-list search covered the current text of EasyList and EasyPrivacy, the uBlock Origin lists, StevenBlack hosts, the AdGuard DNS filter and Peter Lowe’s list. The DNS answers came from two public resolvers, and the registry records from the RDAP service for .xyz. We did not contact Discover-Tech, Candiru or Integrity Labs; the statements attributed to them or their partners come from the cited reports, and Citizen Lab says Discover-Tech and Candiru did not respond to it.

Tags:

Ad TechCandiruCitizen LabSpywareSurveillanceThreat Intelligence

Share

Victorian silver-framed glass pickle jar with a swirled purple and white body and a domed lid against a grey background
Previous Post

How to Stop Pickle Deserialization Attacks in Python and Load Model Files Safely

Iron filings tracing the magnetic field between two bar magnets with their north poles facing each other, a nod to the Magnetar name
Next Post

Microsoft’s Magnetar Turns Hash-Only Proving Into a Laptop Benchmark Win, but Recursion Is Still Unimplemented

No Comment! Be the first one.

Leave a Reply Cancel reply

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

Latest
21 Oct
How to Return RFC 9457 Problem Details From a FastAPI API and Stop Error Responses From Leaking Secrets
21 Oct
Cloudflare’s tsrs Turns an Agent-Written Compiler Port Into a Differential-Testing Bet, and the 8x Headline Depends on the Baseline
Trending
October 21, 2026
How to Return RFC 9457 Problem Details From a FastAPI API and Stop Error Responses From Leaking Secrets
October 21, 2026
Cloudflare’s tsrs Turns an Agent-Written Compiler Port Into a Differential-Testing Bet, and the 8x Headline Depends on the Baseline
October 21, 2026
PHP Warns of Two Critical Flaws and Sets a 14:00 UTC Patch, but the Leak, the Score and the Exploitation Reports Do Not Agree
October 20, 2026
How to Build a Safe AI Agent Harness in Python With Tool Validation, a File Sandbox, and Run Limits
October 20, 2026
Ai2’s FairStress Turns Fairness Testing Into a Pressure Test, but the Released Files Do Not Match the Report
October 20, 2026
An Anonymous Comment Can Hand Duplicator a Payload That Fires at Migration Time, and Under a Third of Free Installs Run the Fix

Related Posts

Rows of server racks in a data center representing network infrastructure targeted by botnets
News

C0XMO Botnet Shows Why Old Router Firmware Still Matters

June 7, 2026
Close-up of a USB flash drive, representing physical data-theft risk in office security incidents
News

Fake IT Support Is Now Walking Through the Front Door

June 7, 2026
A phone security app on a smartphone resting on a laptop keyboard.
News

Everest Forms Pro Flaw Is Being Exploited to Create Rogue WordPress Admins

June 7, 2026
A phone secured by a padlock, illustrating AI data-leak containment and security controls.
News

OpenAI’s Lockdown Mode Is a Data-Leak Brake, Not a Prompt-Injection Cure

June 8, 2026
SXZ.io SXZ.io
  • [email protected]

Categories

Articles
Learning Hub
News

All Rights Reserved by SXZ.io ©2026