Top Highlights
- Cybercriminals are leveraging stolen API keys and session tokens, including unexpired JWTs, to bypass authentication and gain unauthorized access to AI services, often leading to account hijacking and account-based fraud.
- Attackers are selling stolen AI account credentials and tokens on underground forums, enabling activities like LLMjacking, resource theft, espionage, and automated account abuse.
- The use of anti-detect browsers and automation tools allows threat actors to conceal their activities, bypass security controls, and conduct session replay attacks, complicating detection and mitigation efforts.
Threat, Attack Techniques, and Targets
Cybercriminals are exploiting information stealer logs to hijack AI user accounts. They use tools like Lumma Stealer and Vidar to collect data from infected computers. These logs contain credentials, session tokens, and API keys. Threat actors buy these logs on underground forums. They then use the stolen data to attack services from providers like Google, Microsoft, Anthropic, and others. These stolen session tokens and API keys can be reused to bypass multi-factor authentication (MFA) and log into accounts without traditional login. Researchers found thousands of unexpired tokens, including JSON Web Tokens (JWTs), linked to AI services such as OpenAI and others. The stolen tokens include personally identifiable information (PII), which can be used for social engineering. Attackers can also access API keys for AI services like Google Gemini and OpenRouter. They use these keys for espionage, extortion, or resource theft. Criminals may sell access to AI accounts, making the attacks more widespread. Techniques like anti-detect browsers help criminals hide their activities and pass security checks.
Impact, Security Implications, and Remediation
The stolen tokens can give attackers full access to AI accounts, bypassing login and MFA. If these tokens are replayed, attackers can control accounts or use AI resources for malicious activities. The theft of PII is also concerning because it can lead to social engineering and phishing attacks. The use of stolen API keys might enable AI-related resource theft or “LLMjacking,” where criminals use AI resources without permission. These actions can lead to high costs and data theft for organizations. Security measures like IP allowlisting and Device Bound Session Credentials (DBSC) can help prevent token replay attacks. Organizations should secure their API keys, monitor for token reuse, and implement short-lived tokens. If you suspect a breach, get guidance from the relevant vendor or authority to apply appropriate security measures.
Discover More Technology Insights
Explore the future of technology with our detailed insights on Artificial Intelligence.
Discover archived knowledge and digital history on the Internet Archive.
ThreatIntel-V1
