Connecting online dots. Stolen iPhone scam. Influencing AI.
Surveillance through advertising. Using sound to track website users. And why consumer spy detectors don't work.
Only connect
The arrests of two men believed to be members of a prolific cybercrime group are a perfect case study in how a single online account can unravel digital anonymity. Australian police detained the men following a joint investigation with the FBI. They’re alleged to be members of TeamPCP, which has been blamed for attacks on the software supply chain affecting more than 1,000 organisations around the world. The costs of fixing the resulting issues are estimated in the hundreds of millions of dollars.
Independent investigations by Flare and renowned cyber journalist Brian Krebs show how Telegram activity, reused aliases, accounts and other online traces linked the alleged TeamPCP members to real-world identities. In one case, Flare found a distinctive username associated with TeamPCP and used an open source intelligence tool to search for it across hundreds of online platforms. The results included hits that led to a real name, with cross references suggesting the individual was most likely TeamPCP’s de facto leader.
While this story relates to cybercrime, exactly the same techniques can be deployed against anyone with online accounts. That’s why a lot of our work involves carrying out open source searches to uncover what details can be found about journalists working on sensitive projects. This week, the US Navy told sailors, reserves, and civilian staff - and their families - to remove any mention of their ties from social media. The Navy’s concern is that online details could be used to trace military movements or to plan physical attacks. Minimising online personal information is good practice for anyone, but it’s particularly important when working on anything sensitive.
Threats
Lost: If your iPhone’s stolen, beware of claims it has been found. SOCRadar has uncovered a sophisticated operation designed to defeat Apple’s Activation Lock. Armed with details from the stolen device, scumbags bombard owners with convincing emails, texts, WhatsApp messages and even AI-generated calls impersonating Apple Support. The aim is to steal their passcode, Apple ID and two-factor authentication code.
Persistent: A new phishing toolkit tricks victims into allowing attackers to register a passkey on their account, creating access that persists even if the password is subsequently changed. The kit, which is being sold for $10,000, doesn’t mean passkeys are insecure but does mean it’s important to ensure controls are in place to identify and block fake authentication requests. Abnormal uncovered the kit and has good advice on mitigating the threat.
Locked: A fiendish tactic tries to steal a user’s system login credentials by creating a fake Windows lock screen. Expel identified the method while investigating an incident that began with a fake Teams message impersonating the organisation’s IT Helpdesk. It managed to persuade the victim to run an installer presenting itself as “PowerShell Cleaner.” Needless to say, it’s never a good idea to install anything without checking first, even if the instruction seems to come from IT.
Banking: Researchers have identified a $25 template being used to create more than 800 fake bank websites. The sites provide convincing dashboards, as well as transfer interfaces involving completely fictitious financial institutions. The aim appears to be to create online banks that can be used in investment scams and similar frauds. No matter how professional it looks, a website means nothing. Allure
GTA: Fans have now been waiting more than a decade for the release of Grand Theft Auto VI. No surprise then that a rash of fake websites have appeared, impersonating the game’s developer and offering playable demos. Of course they’re neither demos nor playable, but instead are designed to harvest information from any device on which they are installed. Malwarebytes
AI
Influence: Just as companies spent years trying to manipulate Google rankings through search engine optimisation, so political operators are now producing content designed to influence AI output. 404 Media has uncovered a “US think tank” that’s actually funded by the Israeli government and which has published more than 100 articles in the month since its launch. Last month, reports emerged of a $46.5 million contract funded by the Israeli government to promote Israeli narratives and influence AI platforms including ChatGPT, Claude, Gemini, and Perplexity. Meanwhile, OpenAI says it has disrupted a new Russian covert influence campaign.
Switch: The UK’s National Cyber Security Centre (NCSC) has set out basic security steps for organisations involved in designing and operating agentic artificial intelligence (AI) environments. “Agentic AI systems…can automate complex workflows, reduce routine effort and enable people to focus on higher-value tasks,” the NCSC says. However, those same abilities make them potentially dangerous, as amply demonstrated by recent incidents. (In its latest report on an incident last month, OpenAI says nearly 700 agents coordinated as they tried to perform an impossible task). Among the measures recommended by the NCSC is a “kill switch” which allows the plug to be pulled if things get out of hand.
Sharing: Anthropic has switched memory on by default for consumer users of Claude, so details from conversations are retained and reused automatically. That memory is now shared with Cowork, its AI agent that carries out tasks on a user’s behalf - and it runs both ways, with anything Cowork learns flowing back into chat. This means they’re effectively a single system, and Anthropic says the only way to keep them separate is to use different accounts. Subjects such as health and politics are excluded unless you switch them on. There’s an obvious tradeoff here between convenience and privacy, and for now our position remains to lean towards privacy.
Ads: If you use an AI chatbot, have you asked it yet what it knows about you? The result is likely to be…exhaustive. OpenAI says it has put restrictions around the use of conversation data to drive advertising, but it’sexpanding the countries in which it serves up adverts. Following the US and Europe, it will now start showing them in India on its ChatGPT Free and Go subscription tiers. In February, OpenAI said it had more than 100 million weekly active ChatGPT users in India, making it one of the company’s largest markets.
In brief
Surveillance: Journalists are at risk at home and in exile from the machinery that enables targeted advertising. The warning comes from an investigation by the Committee to Protect Journalists which shows how location details and other data harvested by apps and the advertising ecosystem can be bought and used to identify individuals, track their movements and potentially expose confidential sources. To illustrate the issue, the CPJ cites the case of a Belgian journalist whose previous travel was revealed by the data in a free sample obtained from a US data broker. It contained 100 million location records from around one million devices collected over just two weeks. The best defence against tracking is to starve it of data: minimise the data available: restrict app location permissions, reset or disable advertising IDs, avoid unnecessary apps, and don’t take your phone when meeting particularly sensitive sources.
Fingerprints: On a similar theme, Chinese online retailer AliExpress has been caught using inaudible sounds to track visitors to its website. A security researcher identified what was going on after his Bluetooth headphones stopped working when he visited the site. Websites use a panoply of techniques to “fingerprint” visitors - and in some cases to track them across the web. A new utility called Glassbox will show you just how extensive that tracking is, and suggest some ways to control it.
Breaches: Almost nine million people have been impacted by a data breach at three UK airports owned by the Manchester Airports Group (MAG). It says contact details, vehicle registrations and postcodes were stolen. Many of those affected will have signed up to public WiFi at Manchester, East Midlands, and Stansted airports. We recommend having a separate email that is only used for signing up to such services - though it’s better to avoid public WiFi and use mobile data whenever possible. BBC
Rude: You’re probably familiar with those wrong number texts reminding you about meeting up tomorrow or thanking you for lunch yesterday. Being well brought up, you might be tempted to reply. Don’t! Wrong number texts can be the beginning of complex investment scams, or simply a way to confirm a number is genuine. Either way, better to be rude and ignore them. Malwarebytes has a good run through of how they work.
Russia: Three Russian espionage clusters are all focussing their latest tactics on authentication abuse. Google says the “persistent, adaptive phishing campaigns” are using sophisticated social engineering tactics to compromise personal accounts across multiple platforms. It warns that because they’re abusing legitimate authentication, they may not immediately seem like phishing attempts. Among the targets are academics, defence contractors, diplomats and think tanks across Europe and the US.
Reverse: Apple has decided that gutting its Hide My Email feature was a bad idea, so it has made an unusual U-turn. The change would have made it much easier for services to identify use of Hide My Email, effectively rendering it pointless. “After reviewing community feedback,” it will return to the status quo ante.
Miss: Consumer detectors sold to help find hidden surveillance devices are often too difficult to use and interpret, meaning they fail in nearly 60% of cases. University College London researchers asked 34 participants to search for hidden devices in mock living rooms. Those using detectors did only marginally better than those who didn’t. “Many of these products present users with signal-strength readings, technical terminology and radio-frequency measurements that require specialist knowledge to understand,” the researchers found.
Updates
WhatsApp: New security features include additional context for calls from unknown callers and stronger two-step verification which will now require a full password rather than a six-digit code.
Microsoft: This month’s security updates have caused problems with printing and PDF export in some applications. Microsoft says it’s working on a solution and in the meantime has suggested a workaround. Unfortunately, the workaround disables the protections added in the security updates. Bleeping Computer
Zimbra: A serious security issue that was addressed by an update in July is being actively exploited. The US cybersecurity agency has given US government agencies three days to ensure the vulnerability has been patched.
Ubiquiti: Fixes to address three maximum-severity vulnerabilities that can be exploited remotely and bypass authentication.