Tagged articles
data leakage
38 articles · Page 1 of 1
The author shares 15 practical pitfalls encountered while training JEV decision models on Qwen3.5, covering pointer head design, bidirectional attention, data loading, label-sorted collapse, shortcut learning, catastrophic forgetting, joint classification-extraction trade-offs, extraction threshold calibration, and engineering bugs like padding leaks and NCCL deadlocks.
Catastrophic ForgettingGGUFJEV decision model
0 likes · 43 min read

This article explains how Alibaba Cloud's AgentLoop platform continuously audits Dify AI applications by tracing model and tool calls, detecting sensitive data exposure, indirect prompt injection, and privilege misuse, and enabling custom rules to pinpoint root causes in workflow nodes for effective remediation.
AI auditingAgentLoopDify
0 likes · 21 min read

This article outlines a rigorous engineering pipeline for turning human feedback into verified AI improvements, covering fact verification, six-layer root-cause diagnosis, experiment cards with falsifiable hypotheses, segregated test sets to prevent data leakage, and controlled releases with rollback plans, illustrated via video inspection and security alert case studies.
AI EngineeringRoot Cause Analysiscontinuous improvement
0 likes · 29 min read

This article explains why random row-based splitting creates false generalization in enterprise ML, detailing how to define prediction timepoints, combine business entity grouping with time windows, and implement five-layer leakage checks to ensure test data truly simulates unseen production data.
ML evaluationbusiness groupingdata leakage
0 likes · 23 min read

The article analyzes four post‑exam security threats—platform data leaks, phishing attacks, fake hacker‑grade‑change scams, and privacy‑risk social‑media posts—and provides concrete steps for students and parents to protect personal information.
Gaokaodata leakageinformation security
0 likes · 10 min read

OpenAI’s new Lockdown Mode disables network access and external actions in ChatGPT to block prompt‑injection attacks that could exfiltrate private information, trading off features like real‑time browsing, Deep Research, and Agent tasks, and is best used only for handling confidential documents.
AI securityChatGPTLockdown Mode
0 likes · 8 min read

The article dissects three critical Instagram vulnerabilities—AI‑driven account hijacking, mass‑report‑based bans, and massive data leaks—and reveals the step‑by‑step monetisation tactics hackers use, from selling compromised accounts to extorting users and trading personal data on underground markets.
Instagramaccount hijackingcybercrime
0 likes · 8 min read

Security researchers uncovered more than 200,000 documents—including cloud access keys, SSH keys, tax forms and bank statements—leaked from JSONFormatter.org and CodeBeautify.org over seven years, accessible via predictable unauthenticated URLs, and demonstrated that attackers can exploit such data within 48 hours.
JSON formatterSSH-Keyscloud keys
0 likes · 8 min read

Researchers built a 13‑billion‑parameter model trained only on texts published before 1931, called Talkie‑1930, and used surprise‑based metrics, programming tests, and a modern‑twin comparison to explore how far such a historically‑constrained model can extrapolate future knowledge and reveal data‑leakage challenges.
AI researchHumanEvaldata leakage
0 likes · 10 min read

Recent data‑leak incidents at AI coding platforms such as Lovable, Vercel, and Anthropic reveal that insecure default settings, supply‑chain compromises, and inadequate threat modeling expose user code and chat logs, prompting experts to warn about the trade‑off between usability and security.
AI programmingAnthropicLovable
0 likes · 7 min read

The article dissects DeepSeek’s series of security breaches in early 2025—including an open ClickHouse database, multiple XSS flaws, model‑level attacks, and regulatory fallout—highlighting how rapid AI product rollout can outpace essential security safeguards.
AI securityClickHouse exposureDeepSeek
0 likes · 14 min read

This article analyzes the systemic risks of QR code package leakage in fast‑moving consumer goods marketing, breaks down the four‑stage code lifecycle, identifies three critical vulnerability points, and presents a five‑layer security framework—including authorization, encryption, expiration, real‑time risk monitoring, and full‑traceability—to protect digital identity assets.
FMCGIdentity PlatformQR code security
0 likes · 12 min read


SuanNiMar 25, 2026 · Artificial Intelligence
This article explains the evaluation pillar of context engineering, introduces the three core RAG metrics (context relevance, faithfulness, answer relevance), details the RAGAS automated assessment framework, shows how to build evaluation datasets, adopt evaluation‑driven development, and protect RAG systems from prompt injection and data leakage.
LLMPrompt InjectionRAG
0 likes · 13 min read

A leading identity verification provider exposed over one billion personal records after a cloud storage bucket was misconfigured, revealing names, IDs, biometric data and more; the breach impacted finance, e‑commerce, government and social platforms, prompting analysis of technical and managerial failures and a set of remediation steps for individuals, enterprises and the industry.
KYC securitycloud misconfigurationdata leakage
0 likes · 10 min read

The OpenClaw AI agent, nicknamed “lobster,” has sparked a nationwide frenzy across all ages, but security monitoring reveals over 40,000 public instances, 63% vulnerable and 12,000 fully controllable, making it a prime target for network attacks and data leakage.
AI agentGitNode.js
0 likes · 6 min read

A Shenzhen tech employee was fired after his girlfriend posted his badge and 3‑million‑yuan salary online, sparking a debate about strict corporate data protection, internal reporting mechanisms, and the broader responsibilities of both companies and individuals in safeguarding sensitive information.
Corporate policydata leakageemployee privacy
0 likes · 4 min read


21CTOMar 10, 2025 · Information Security
Microsoft uncovered a malicious ad campaign that used pirated streaming sites and multi‑layer redirects to deliver GitHub‑hosted malware, while a separate Surfshark report revealed popular phone‑cleaning apps secretly sharing extensive user data with hundreds of third‑party partners.
Ad FraudGitHubdata leakage
0 likes · 4 min read

A recent case where a maintenance worker exploited device‑management flaws to steal confidential files for foreign spies highlights the need for heightened vigilance, strict self‑discipline, and prompt reporting, offering practical steps to safeguard against similar security breaches.
data leakageincident responseinformation security
0 likes · 4 min read

Researchers from Truffle Security revealed that GitHub's delete function often fails to truly remove data, exposing a new Cross Fork Object Reference (CFOR) vulnerability that allows anyone with a commit hash to access deleted or private repository data, posing serious security risks.
CFORGitHubSecurity
0 likes · 11 min read

The article details a security case where an unauthenticated university API allowed an attacker to enumerate and download thousands of students' personal data by manipulating pagination parameters after discovering default credentials from a leaked PDF.
API enumerationStudent Datadata leakage
0 likes · 7 min read

This article examines recommendation system problem definitions, differences between academia and industry, offline evaluation pitfalls and data leakage issues, data construction challenges with datasets like MovieLens, and evaluates whether large language models can serve as effective solutions for modern recommendation tasks.
Offline Evaluationdata leakagelarge language models
0 likes · 20 min read

This article examines how large language models fit into recommendation systems by discussing problem definitions, offline evaluation pitfalls such as data leakage, dataset construction issues exemplified by MovieLens, and the practical limits of using LLMs as a universal solution.
MovieLensOffline Evaluationdata leakage
0 likes · 18 min read

This article presents a comprehensive overview of digital watermarking, covering its history, evaluation metrics, various media implementations, challenges posed by AI, and practical applications—especially in e‑commerce—to protect data throughout its lifecycle and enable effective leakage tracing.
data leakagedatabase watermarkdigital watermarking
0 likes · 16 min read

A CCTV investigation revealed that some mobile browsers and integrated third‑party SDKs silently read users’ clipboard, capturing bank account numbers, passwords and other personal data in plaintext even when running in the background, highlighting serious privacy risks under China’s Personal Information Protection Law.
CCTVclipboarddata leakage
0 likes · 3 min read

This article examines the rapid growth of China’s digital economy, the escalating risk of data leaks, and how digital watermarking—across images, text, and databases—can be applied to trace leaks, protect privacy, and address practical challenges in e‑commerce environments.
data leakagedatabase watermarkdigital watermarking
0 likes · 15 min read

The article explores the rapid growth of China's digital economy, the escalating risk of data leaks, and how digital watermarking—across images, text, and databases—can be employed to trace leakage sources, protect e‑commerce data, and address practical challenges in security implementations.
data leakagedatabase watermarkdigital watermarking
0 likes · 15 min read

The author examines the AAAI 2022 paper “Mind Your Clever Neighbours,” reveals that it exploits a data‑leak in unsupervised person re‑identification, critiques the unnecessary Graph Correlation Learning step, and demonstrates a much simpler averaging method that yields superior results.
Unsupervised Learningdata leakagegraph correlation learning
0 likes · 6 min read

In today's big‑data era, excessive stress leads many to seek relief through risky online activities, but unauthorized app permissions and visits to dubious sites can expose personal information, so users must stay vigilant, limit permissions, avoid harmful sites, and use security tools to protect their mobile privacy.
big datadata leakagemobile security
0 likes · 6 min read

Even when you use a browser’s incognito or private mode, your search queries, visited sites, and personal data can still be captured by browsers, cookies, your employer, ISP, and even input methods, making true online anonymity virtually impossible.
browser trackingcookiesdata leakage
0 likes · 4 min read

GitHub Copilot unexpectedly generated a fabricated ID for Bilibili CEO Chen Rui, sparking concerns about how large language models trained on public data can inadvertently expose synthetic personal information and highlighting broader privacy and ethical issues in AI‑driven code assistants.
AI privacyGitHub Copilotcode generation
0 likes · 6 min read

Facial recognition technology, now common in payments, transport and law enforcement, often captures more than just the face, uploading full‑camera images to back‑ends where human reviewers may see users in vulnerable situations, raising serious privacy and security concerns.
AI ethicsdata leakagefacial recognition
0 likes · 7 min read

A recent exposé reveals that many facial‑recognition systems capture the entire camera view and send images to backend staff for manual review, exposing users—often in vulnerable moments—to privacy breaches and sparking widespread concern across social media.
app securitydata leakagefacial recognition
0 likes · 6 min read


21CTOApr 24, 2020 · Artificial Intelligence
This article examines the frequent pitfalls of recommendation systems—misleading metrics, over‑optimizing precision, data leakage, feature inconsistencies, and distribution bias—that cause offline AUC improvements to translate into lower online CTR and CPM, and offers practical mitigation strategies.
AIExploitationMetrics
0 likes · 15 min read

The article surveys typical challenges in recommendation systems, including ambiguous evaluation metrics, the trade‑off between precise algorithms and user experience, the exploration‑exploitation dilemma, and why offline AUC improvements often lead to online CTR/CPM drops due to data leakage, feature inconsistency, and distribution shifts.
AUCCTRExploration-Exploitation
0 likes · 14 min read

The popular Windows file‑search tool Everything includes an HTTP server that, when left unsecured, lets anyone who knows your IP and port browse and download every file on your computer, exposing personal data that can even be indexed by search engines, but simple configuration changes can close the leak.
HTTP serverdata leakagefile search
0 likes · 7 min read

This guide explains recent cloud‑based data‑leak incidents, categorizes common leak vectors, analyzes technical and managerial root causes, and provides actionable monitoring techniques, rule‑configuration examples, and incident‑response steps using Tencent Cloud Security Operations Center.
GitHubTencent Cloudcloud security
0 likes · 19 min read

In this interview, a solo female participant from Shanghai Jiao Tong University shares her experience, challenges, and technical insights—including feature engineering, memory management, and model tuning with XGBoost and LightGBM—gained while competing in the Tencent Social Ads algorithm contest.
Model tuningTencentXGBoost
0 likes · 5 min read

The article explains how various companies use technologies such as ultrasonic audio, facial recognition, smart TV monitoring, and image analysis to collect personal data, outlines the common pathways of information leakage, and offers practical advice for safeguarding online privacy.
data leakageinformation securitypersonal data
0 likes · 10 min read
