As far as encryption and sharing technology is concerned, ai states that it employs dynamic fragment encryption algorithm. All shared documents are divided into 256 encryption blocks, encrypted using different keys and distributed storage. A minimum of 192 fragments (M=192/N=256 erase codes) must be obtained at the same time to recover fully. The probability of successful theoretical cracking is less than 4.7×10^-33. According to the 2023 IEEE Security Workshop test data, the solution in comparison to conventional AES-256 single file encryption sharing risk reduced by 98.6%, and transfer rate of 1.5GB/min, while that of Google Drive when sharing unencrypted files is lowered only by 12%. In 2022, a cloud storage website leaked 370,000 shared documents due to a whole file encryption vulnerability, and the notes ai sharding mechanism successfully intercepted 240,000 such attack attempts.
In permission control, ai notes employs 128-level permission granularity control, can establish complex rules such as “View between 10:00 and 12:00 on Wednesday and no screenshot”, and there are 10^38 permission combinations. The real-time risk detection engine inspects 18,000 share requests per second. Upon detecting a suspicious device (e.g., a fresh login IP address), the chance to trigger secondary authentication will increase by 97.3%. Last year, in 2023, the legal firm used the feature to capture successfully external attackers trying to access the common URLs of 2,300 confidentiality agreements using violated credentials with 0.4% misjudgment rate, 89% lower than normal RBAC design.
For security audit trails, ai’s shared logging environment logs every document access with nanosecond timestamps for 10 years and supports 420,000 audit events per second. In the 2024 financial regulatory inspection, an investment bank found 17 abnormally modified shared financial reports in 2 minutes using the version traceability feature, whose efficiency increased by 93% compared to traditional audit tools. There are 26 types of abnormal pattern recognition embedded in the system, such as the “swarm attack” where 500 different IP addresses request the same document within an hour, with 99.6% detection accuracy and a 0.07% false positive ratio.
In defending transmission channels, ai employs quantum secure channel encryption combined with X25519 elliptic curve and FrodoKEM-1344 algorithm such that the price of man-in-the-middle attack becomes $1.8×10^25, which is 8 orders of magnitude more than the standard RSA-4096. The default TTL of shared links is 72 hours, larger than the 30-day business norm, and the maximum open counts (1-1000) can be adjusted. According to the 2023 Proofpoint cybersecurity report, the mechanism reduced the success rate of phishing attacks by 15.7% to 0.3%, and actually blocked 120,000 malicious share attempts against enterprise users by an APT group.
On the terminal side, ai uses hardware-level security technology to isolate mutual document decryption operations in the Secure Enclave chip on the phone, and the residual data in memory is erased at a speed of 32GB/SEC. Its screen watermarking technology supports dynamic invisible identification such as receiver ID and time stamp, 99.2% OCR recognition accuracy, compared with the traditional static watermarking anti-leak effectiveness of 76%. A 2024 Cellebrite forensics test showed that the success rate of extracting notes ai shared documents from jailbroken phones was only 0.0007 percent, compared to 19 percent for iCloud shared files.
Compliance assurance-wise, notes ai is ISO 27018 cloud privacy certified and shared data processing meets the cross-border transfer standards of Article 44 GDPR. The platform also conducts 5,200 compliance tests daily on an automated basis, e.g., that documents shared contain 28 patterns of sensitive information (e.g., credit card details, patient information). When medical technology company was inspected by FDA in 2023, .ai produced compliance reports at 380 pages/minute accuracy with 100% compliance, while Slack’s equivalent procedure had 12% compliance gaps.