Exposed: How DeepSeek Tracks You and Sends Data to ByteDance
Security Analysis of DeepSeek iOS App: Encryption Flaws, Privacy Risks, and Data Exposure
Table of contents
- 1. Insecure Symmetric Encryption with Hardcoded Keys
- 2. Insecure Data Storage: Username, Password, and Encryption Keys
- 3. Data Collection and Fingerprinting Risks
- 4. Data Transmission to ByteDance's Volcengine
- 5. Third-Party Data Sharing: Intercom SDK
- 6. Missing iOS Security Controls
- 7. Network Traffic and Chinese Data Access Concerns
- Conclusion: Why This Matters & Next Steps
Mobile applications play a crucial role in today’s digital landscape, handling vast amounts of sensitive user data. However, improper security implementations can expose users to data breaches, privacy violations, and cyber threats. This technical blog analyzes multiple security flaws in the DeepSeek iOS app, focusing on encryption vulnerabilities, insecure data storage, fingerprinting risks, and network traffic concerns.
1. Insecure Symmetric Encryption with Hardcoded Keys
Encryption is Only as Strong as Its Implementation
Encryption is a fundamental mechanism to protect data confidentiality and integrity. However, improper implementation can render it ineffective. The DeepSeek iOS app exhibits several weaknesses in its encryption approach:
Use of a broken encryption algorithm (3DES): The app employs Triple Data Encryption Standard (3DES), which is deprecated due to known vulnerabilities and susceptible to brute-force attacks.
Hardcoded encryption key: The encryption key is hardcoded within the app, making it easily retrievable through reverse engineering.
Use of NIL for Initialization Vector (IV): The absence of a random IV reduces the encryption strength, making ciphertext predictable.
Reuse of IV: If an IV is reused across multiple encryption instances, it significantly weakens security by allowing attackers to correlate encrypted data.
Reverse Engineering Analysis
Using r2ai (an AI-enhanced reverse engineering tool for radare2) and Frida (a binary instrumentation framework), security researchers analyzed the encryption function in the BDAutoTrackLocalConfigService
service, particularly the saveUser
call.
Frida Hook Output of Encryption Function
plaintextCopyEdit================ CCCrypt Call ================
Operation => kCCDecrypt
Algorithm => kCCAlgorithm3DES
Options => kCCOptionPKCS7Padding
Key size => 24
Key => MEMyNyRTU2
IV => None
Data input => 4pvOAF6luXJQ==
Data output => NjRjOWxNmFk
This confirms the use of an insecure symmetric encryption algorithm with a hardcoded key and missing IV, making the encryption ineffective against real-world attacks.
2. Insecure Data Storage: Username, Password, and Encryption Keys
Cached Sensitive Data in Local Storage
Analysis of the DeepSeek iOS app revealed that user credentials and encryption keys were stored insecurely in a cached database on the device. In scenarios where an attacker gains physical access to an unlocked device, this data can be easily extracted and exploited.
NSURLRequest API Caching Issue
The app uses NSURLRequest API to communicate with remote servers. By default, this API caches HTTP responses in a Cache.db
file unless explicitly disabled by developers. This means sensitive data remains stored locally, creating a serious security risk.
3. Data Collection and Fingerprinting Risks
Fingerprinting: The Silent Privacy Violation
Modern mobile applications track user behavior, device information, and network metadata. While some tracking is necessary for app functionality, excessive data collection can pose serious privacy risks, particularly when combined with external data sources.
How Minimal Data Can Lead to Identity Exposure
Even seemingly harmless data points can reveal a user’s identity when correlated with other datasets. Consider the following scenario:
User 1 asks: "What Elements Are Commonly Used as Nanoparticles?"
- Data suggests they are a student researching nanotechnology.
User 2 asks the same question but:
Uses a high-end iPad
Is connected via FirstNet (a U.S. public safety broadband network)
Uses a cellular data connection
These additional data points may categorize User 2 as a high-value target, increasing the risk of surveillance or cyber espionage.
Data Points Collected by the App
DeepSeek iOS collects extensive user data, including:
Device name (potentially containing real names)
OS version
Language settings
Network connection type (WiFi, Cellular, VPN, etc.)
App usage statistics
Such data can be combined across multiple apps, making it easy to de-anonymize users and track them across different services.
4. Data Transmission to ByteDance's Volcengine
What Is Volcengine?
Volcengine is ByteDance’s cloud service platform, used for AI, analytics, and enterprise solutions. While ByteDance is headquartered in China, its cloud infrastructure spans multiple countries, leading to concerns about data sovereignty and access by foreign governments.
Network Analysis of Data Transmission
The DeepSeek app transmits sensitive user data to apmplus.volces.com/monitor/collect/c/perfor... Key concerns include:
Volcengine’s connection to ByteDance: Data sent here could be accessed under Chinese data laws.
Fingerprinting data (e.g., device name): This allows direct association between users and their devices, reducing anonymity.
5. Third-Party Data Sharing: Intercom SDK
The app integrates Intercom's iOS SDK, which further transmits user data to an external platform. Security researchers identified fingerprinting vectors in this data exchange, such as:
Color scheme tracking
Voice-over feature status (used for accessibility tracking)
Left-to-right (LTR) language setting
These attributes may seem harmless but contribute to building a highly unique user fingerprint, enabling tracking across multiple apps and services.
6. Missing iOS Security Controls
iOS Security Protections Ignored
Apple provides various security features to protect app users, but DeepSeek fails to leverage them properly:
Security Control | Status in DeepSeek iOS App | Risk |
App Transport Security (ATS) | Globally disabled | Allows unencrypted data transmission |
Privacy Manifest Disclosure | Not properly declared | Fingerprinting APIs not disclosed |
Data Encryption Best Practices | Violated | Weak encryption (3DES, hardcoded key, missing IV) |
User Defaults API Usage | Misused for fingerprinting | Risk of user tracking |
7. Network Traffic and Chinese Data Access Concerns
Traffic Routing Through U.S.-Based ByteDance Servers
While DeepSeek app’s network logs show data routing through U.S. IP addresses, reverse analysis reveals ByteDance ownership.
Example: DNS Resolution of ByteDance Server
plaintextCopyEditping apmplus.volces.com
PING apmplus.volces.com.queniusz.com (8.45.52.229)
Geo-located in the U.S.
Owned by Level 3 Communications (acquired by Lumen)
Connected to ByteDance (China-based parent company of TikTok and Volcengine)
Implications for Enterprises and Government Agencies
Organizations using DeepSeek risk exposing:
Intellectual Property (through AI model usage data)
Government or Military Personnel Data (through device fingerprinting)
User Privacy (via metadata tracking and third-party data sharing)
Conclusion: Why This Matters & Next Steps
The DeepSeek iOS app demonstrates critical security flaws that impact user data confidentiality, integrity, and privacy. Key takeaways include:
Weak encryption (3DES, hardcoded key, missing IV) compromises data security.
Insecure storage of sensitive data (credentials, encryption keys) increases attack risks.
Fingerprinting and data tracking pose serious privacy concerns.
Data transmission to ByteDance servers raises geopolitical and regulatory risks.
Recommendations for Developers
To address these risks, developers should:
Use modern encryption standards (AES-256) with dynamic IVs and secure key storage.
Disable automatic caching for sensitive data using NSURLRequest.
Limit data collection and disclose API usage per Apple’s privacy policies.
Implement App Transport Security (ATS) to enforce encrypted data transmission.
Final Thought
Security is not just about compliance; it's about protecting users from real-world threats. Apps like DeepSeek highlight the importance of thorough security testing before deployment. Enterprises and users must remain vigilant, ensuring that the apps they rely on respect both security best practices and user privacy.