Low-tech solutions emerge as best defense against deepfake threats
Experts recommend simple, controlled entry points against AI-powered fraud
1$25 million deepfake scam
In January 2024, an employee at a professional services firm joined a video call with what they believed was the company's CFO. This resulted in 15 wire transfers totaling $25 million. All participants were AI clones created from executives' public appearances.
2Traditional authentication methods fail
Deepak Gupta, Technical CEO at GrackerAI, states 'Seeing and hearing someone is no longer proof they're real.' Face and voice recognition protocols have become ineffective, creating significant risks for companies handling millions in transactions annually.
3Deepfake detection rates decline
A University College London study showed participants could only identify deepfakes with 73% accuracy. Training improved this by just 3.84%. Meta-analysis of 56 studies revealed human detection is now nearly random.
4Conventional detection methods insufficient
A joint notice from NSA, FBI and CISA declared traditional automated detection protocols ineffective. These methods assumed statistically detectable manipulation traces, but new technologies eliminate these markers.
5Social engineering attacks proliferate
In 2024, security training firm KnowBe4 was targeted by a fake employee who passed interviews without suspicion. This North Korean operative installed malware. Experts report North Korea places thousands of fake employees in US and European companies annually.
6US Department of Justice warning
In 2025, the DOJ revealed North Korean attackers collaborate with operatives in the US, China, UAE and Taiwan. Google's Threat Intelligence Group has investigated similar incidents since 2022, noting increasing targeting of European nations.
7Low-tech solutions prove effective
Experts state the most effective defense against advanced AI attacks is often deliberately low-tech. While deepfakes can replicate public appearances, they're useless against systems that exist outside observable channels and can't be remotely monitored.
8Hardware security keys recommended
Security consultants now recommend hardware-based security keys and verbal passphrases, based on government guidance. These methods, relying on single controlled entry points, provide superior protection against sophisticated AI attacks.
Original article: Latest news — Derin sahte tehdidine karşı düşük teknolojili çözümler öne çıkıyor