2026-02-20 (FRIDAY): AI-Accelerated Malicious Chrome Extension Campaigns AUTHOR(S): Nabeel Mohamed, Qinge Xie, Shresta Bellary Seetharam, Fang Liu, Oleksii Alex Starov REFERENCES: - https://www.linkedin.com/posts/unit42_affiliate-genai-ai-activity-7431848606954905601-_ily/ - https://x.com/Unit42_Intel/status/2026082989071798598 NOTES: We keep track of LLM/AI-generated malicious Chrome extensions. Threat actors are increasingly weaponizing LLM/AI code-generation tools to mass-produce malware. We identified two different campaigns that were using code generation to accelerate their attacks.   -- 6 malicious Chrome extensions that are being used in an affiliate hijacking campaign   -- 4 malicious extensions providing dual functionality of GenAI productivity tool and sensitive data exfiltration Attackers are moving at the speed of AI. Our defense must too. DETAILS: These extensions are AI-generated and malicious, affecting users across multiple e-commerce platforms. As mentioned in the below reference, all extensions inject affiliate tracking links into the traffic to shopping sites without explicit user consent. AI-Generated Tell-Tale Signs:   -- Overly verbose explanatory comments: AI language models tend to generate excessive comments that explain obvious operations in unnecessary detail. Examples: aohfjaadlbiifnnajpobdhokecjokhab, gjlbbcimkbncedhofeknicfkhgaocohl, dkhpgbbhlnmjbkihoeniojpkggkabbbl   -- Strategy labeling with section headers: AI models often organize code with elaborate section headers using dashes or numbered steps mimicking tutorial-style code. Example: aohfjaadlbiifnnajpobdhokecjokhab   -- Generic error handling with minimal logging: AI-generated code often includes try-catch blocks that simply log or ignore errors without proper recovery logic. Example: mpgaodghdhmeljgogbeagpbhgdbfofgb   -- Defensive coding patterns: AI models tend to add unnecessary defensive checks that protect against edge cases that cannot realistically occur given the calling context. Example: mpgaodghdhmeljgogbeagpbhgdbfofgb   -- Template-like structure with generic naming: AI-generated code often uses generic, textbook style conventions and structural patterns that feel like they came from documentation examples rather than real-world applications. Example: mcaihdkeijgfhnlfcdehniplmaapadgb, eebihieclccoidddmjcencomodomdoei   -- Unfinished patterns and placeholder comments: AI-generated code sometimes includes comments suggesting future implementations or expressing uncertainty. Example: mpgaodghdhmeljgogbeagpbhgdbfofgb   -- Systematic naming conventions: AI tends to use systematic naming that follows strict patterns across the codebase, lacking the organic variation found in human-written code. Examples: All affiliate hijacking extensions mentioned   -- Identical code patterns across extensions: Multiple extensions share nearly identical code structures for the same functionality, suggesting that they were generated from similar prompts. Example: mcaihdkeijgfhnlfcdehniplmaapadgb, mpgaodghdhmeljgogbeagpbhgdbfofgb, nppjmiadmakeigiagilkfffplihgjlec Some strategies to defend against AI-generated malicious extension threats:   -- Deploy code signatures to spot AI patterns in suspicious extensions   -- AST (Abstract Syntax Tree) analysis to detect comment-to-code ratio, systematic naming pattern compliance   -- Similarity detection across known AI-generated malicious extensions and new extensions INDICATORS: IOCs - Extension IDs for Affiliate Hijacking Campaign: aohfjaadlbiifnnajpobdhokecjokhab gjlbbcimkbncedhofeknicfkhgaocohl hkhmodcdjhcidbcncgmnknjppphcpgmh mcaihdkeijgfhnlfcdehniplmaapadgb mpgaodghdhmeljgogbeagpbhgdbfofgb nppjmiadmakeigiagilkfffplihgjlec IOCs - Malicious GenAI Extension IDs: bonhfflnjgdbnhcpjemkknlhimceckgb dkhpgbbhlnmjbkihoeniojpkggkabbbl eebihieclccoidddmjcencomodomdoei iefpkdilnfhogjbkhgnliaomoldgkdlj REFERENCES: https://socket.dev/blog/malicious-chrome-extension-performs-hidden-affiliate-hijacking