AdFection: targeted ads confirmed as a zero-click spyware delivery channel
Citizen Lab and Confiant document an Israel-based ad-tech firm whose ads profiled devices worldwide and, at Al Jazeera, led to Candiru spyware. The exposure footprint spans government and intergovernmental networks.
The ad slot became the exploit delivery channel
Commercial advertising infrastructure has now been documented as a working delivery path for zero-click mercenary spyware. On 15 October the Citizen Lab and the ad-security firm Confiant published co-timed investigations into Discover-Tech, an Israel-based company that presents itself as an advertising business [1][2]. The Citizen Lab calls the technique AdFection and states that this is “the first-ever public confirmation of zero-click mercenary spyware delivered via targeted ads” [1]. The finding matters less for its novelty than for what it removes. Delivery no longer depends on a lure, a link or a compromised website. It depends on a target’s device rendering an ad.
Ad-based delivery had been described before, but mostly from leaked vendor material. Amnesty International documented Intellexa’s Aladdin vector in December 2025 from internal documents, and framed active use as its belief rather than a confirmed infection [11]. Google Threat Intelligence Group reported in the same week that a few Intellexa customers used malicious advertisements on third-party platforms to fingerprint users and redirect selected targets to exploit servers [10]. The October reports add victim-side network evidence.
What the two reports show
The confirmed victim set is a newsroom. The Citizen Lab reports that at least 45 devices at Al Jazeera were targeted in 2024 and 2025 [1]. Network logs shared by the broadcaster show a sequence: an ad impression reaching Discover-Tech infrastructure, then contact with a Candiru validation server, then a Candiru command-and-control server [1]. The Citizen Lab attributes the payload to Candiru, which it says now operates as Integrity Labs (unrelated to the Chinese firm Integrity Technology Group), with high confidence. It bases this on a Windows implant recovered from a compromised device and on overlaps with Candiru’s 2021 attack chain [1]. Confiant states plainly that it has not captured a full exploit chain in the wild. Its own Candiru attribution rests on the Citizen Lab findings and on tradecraft overlaps with ESET’s 2021 research [2].
Those overlaps have a history. Microsoft reported in 2021 that the actor it tracks as SOURGUM, which the Citizen Lab assessed with high confidence to be Candiru, had more than 100 victims, including in Armenia, Turkey, the UK and Singapore [8]. ESET assessed with medium confidence that operators of 2020–2021 watering-hole attacks on Middle Eastern websites were Candiru customers. Those sites fingerprinted visitors and passed selected ones towards browser exploits [9]. The US Commerce Department added Candiru to the Entity List in November 2021 for supplying spyware to foreign governments [7]. AdFection moves that same select-then-exploit pattern from compromised websites into programmatic bidding.
The scale figures differ between sources, mostly because they measure different things over different windows:
- The Citizen Lab analysed about four billion bidding records from March to June 2026. Discover-Tech’s buying platform bid on about 3.77 billion impressions and won about 442 million. Ads carrying a script tied to the operation appeared in 75,373 apps and on 120,768 websites, with bids geolocated to 161 countries [1]. The Citizen Lab notes that this data excludes two of the buying channels it identified [1].
- Confiant, observing from 1 April to 10 September 2026, estimates about 10 million impressions per day across 228 countries [2]. Reuters reported Confiant’s view that “very likely hundreds of millions” of people were affected [3].
- Characterisation of the confirmed victims also varies. Citizen Lab and CPJ refer to at least 45 devices [1][6]. The Record describes 45 journalists [5]. Al Jazeera’s own report says dozens of devices [4].
The exposure footprint matters more than the confirmed infections. The Citizen Lab found impressions on networks of NATO, the UN, EU institutions, Nordic governments and critical infrastructure, and Australian critical infrastructure and research institutions. It also found them on government-operated networks in around two dozen countries [1]. Confiant reported placements on Pentagon, International Atomic Energy Agency and International Criminal Court networks [3][4], and on Estonia’s Information System Authority [5]. Both researchers caution that an impression on an organisation’s network shows profiling exposure, not infection [1][2].
Two capabilities, not one
The reports describe a targeting layer as well as a delivery layer. Confiant observed ad creatives reading the page’s cookies and sending them, with browser client hints, to Discover-Tech domains. It observed this across more than 1,000 publisher domains, including where consent for processing was absent [2]. Confiant assesses that the resulting identity graph could let a spyware customer select deanonymised targets by advertising ID [2]. The Citizen Lab found that many websites sync identifiers with Discover-Tech domains even when no Discover-Tech ad is served [1]. It also describes a separate actor, which it names DARK CARTOGRAPHER, using the same infrastructure to link advertising IDs to phone numbers. That actor abuses mobile-carrier weaknesses in more than 30 countries, and a fuller report is to follow [1].
Confiant’s regional data shows that APAC exposure is not hypothetical. It recorded peak reach of 0.62% of impressions in Singapore and 0.61% in Armenia [2]. Both countries also appeared in Microsoft’s 2021 Candiru victim list [8]. Peak reach is not targeting, and neither report names an APAC victim.
Platform response is partial
Google removed Discover-Tech from its ad platforms for systematically violating its policies. IAB Europe suspended it from its approved-vendor list, and Comcast’s Beeswax removed it [3]. The Citizen Lab also lists action by Vungle (Liftoff) and Eskimi, and notes that other vendors had not responded by publication [1]. Yandex did not respond to Reuters [3]. Neither Discover-Tech nor Candiru commented [3][6]. Removing one buyer does not remove the technique. GTIG reported that Intellexa had created companies to enter the ad ecosystem, and partners shut those accounts down [10]. Amnesty states that other mercenary vendors and some governments are developing ad-based methods [11].
What defenders should do
Treat advertising as untrusted executable content arriving at privileged endpoints. Practical measures follow from the reports:
- Block ad delivery on high-risk endpoints. Citizen Lab recommends ad-blocking as a cybersecurity practice and warns there is currently no consistently safe way to be exposed to targeted ads [6]. Enforce this through DNS or proxy filtering on managed fleets rather than relying on individual browser extensions.
- Disable advertising identifiers on managed mobile and Windows devices. The US Air Force, Army, Navy and Special Operations Command have reported doing so [12]. Senator Wyden has said these measures have not neutralised the location-data threat [12]. Treat this as one control among several.
- Prioritise browser and operating-system patching for journalists, diplomats and senior officials. Ad delivery replaces the lure, but the payload still needs an exploit chain. GTIG’s guidance remains rapid patching [10].
- Add ad-tech egress to threat hunting. Unexpected repeated contact from sensitive hosts to little-known demand-side platforms or ID-sync domains is a detection category worth baselining, without depending on any single indicator list.
- Make supply-chain vetting a requirement in procurement. Citizen Lab says ad vendors must improve threat detection and audit partners across their supply chains [5]. Publishers serving government audiences can ask the same of their ad partners.
Close
For cyber operations, AdFection describes access as a purchasable commodity. A spyware customer bids on a profiled audience, and the ad network handles distribution. Wolfie Christl, a senior researcher at the Citizen Lab, told Reuters that online advertising as it works today is “a systemic security threat” [3]. The evidence base is still narrow: one confirmed victim organisation, and one operator that has not responded. The infrastructure it ran on is ordinary programmatic advertising, and that infrastructure remains in place.
Sources
Every R3KONX article cites its primary material. 12 sources, in order of first citation. Links open the original publication.
- Exposing AdFection: How Discover-Tech Turned Targeted Ads Into a Pipeline for Zero-Click Mercenary Spyware The Citizen Lab · 2026-10-15
- ADINT: Weaponized Ad Tech for Zero-Click Spyware and Mass Identity Vacuuming Confiant · 2026-10-15
- Israeli firm used online ads to run vast spy operation, researchers say Reuters (via BNN Bloomberg) · 2026-10-15
- Israeli ad-tech firm helped place spyware on Al Jazeera devices Al Jazeera · 2026-10-15
- Zero-click spyware infecting devices through online ads, researchers say The Record by Recorded Future News · 2026-10-15
- Report: Israeli ad-tech firm enables spyware targeting of at least 45 Al Jazeera devices Committee to Protect Journalists · 2026-10-15
- Commerce adds NSO Group and other foreign companies to Entity List for malicious cyber activities US Department of Commerce, Bureau of Industry and Security · 2021-11-04
- Protecting customers from a private-sector offensive actor using 0-day exploits and DevilsTongue malware Microsoft Threat Intelligence · 2021-07-15
- Strategic web compromises in the Middle East with a pinch of Candiru ESET Research (WeLiveSecurity) · 2021-11-16
- Intellexa's prolific zero-day exploits continue Google Threat Intelligence Group · 2025-12-03
- To Catch a Predator: Leak exposes the internal operations of Intellexa's mercenary spyware Amnesty International Security Lab · accessed 2026-10-18
- Exclusive: US military turns off ad trackers on devices amid Middle East targeting reports Reuters (via The Star) · 2026-09-04
Researched and written by the R3KONX analysis desk from the cited primary material. Methodological caveat: scale figures come from two research teams using different datasets and time windows, ad impressions on an organisation's network indicate exposure rather than infection, and the Candiru attribution rests on the Citizen Lab's assessment, with Confiant not capturing a full exploit chain. Corrections to event@r3konx.asia.
