# pii

Personally identifiable information (PII) is information that can distinguish or trace an individual's identity, alone or combined with linked information.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Laranon: Reversible PII Anonymization for Laravel and LLMs

DevFeed: [Laranon: Reversible PII Anonymization for Laravel and LLMs](<https://devfeed.tech/articles/laranon-reversible-pii-anonymization-for-laravel-and-llms-147031.md>)

Original publisher: [Read original article](<https://laravel-news.com/laranon-pii-anonymization>)

Author: Yannick Lyn Fatt

Published: 2026-11-20T01:44:50Z

Content type: article

Language: en

Sources: [Laravel](<https://devfeed.tech/sources/laravel.md>)

Topics: [pii](<https://devfeed.tech/topics/pii.md>), [Laravel](<https://devfeed.tech/topics/laravel.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [laravel](<https://devfeed.tech/tags/laravel.md>), [laravel-packages](<https://devfeed.tech/tags/laravel-packages.md>), [llms](<https://devfeed.tech/tags/llms.md>), [packages](<https://devfeed.tech/tags/packages.md>), [personal-data](<https://devfeed.tech/tags/personal-data.md>), [pii](<https://devfeed.tech/tags/pii.md>), [tokens](<https://devfeed.tech/tags/tokens.md>)

### AI overview

Laranon detects personal data in Laravel text, replaces it with tokens before an LLM call, and restores values from a server-side map. Its excerpt describes in-memory sessions for one request and encrypted cache or database scopes for work spanning requests; scopes expire after six hours by default and can be deleted with `forget()`. It supports token, reversible Faker, and irreversible redaction strategies, plus English and Spanish detection packs, checksum validation for specified identifiers, streaming restoration, log and HTTP scrubbing, and a scan command that reports matches and confidence. Name detection tokenizes individual words and relies on title and name lists, so it does not identify people and may leave names unmatched. Nearby-word context is required for postal codes, ZIP codes, and passports. The package requires PHP 8.2+ and Laravel 12+.

### Source excerpt

Laranon finds personal data in Laravel apps and replaces it with tokens before it reaches an LLM, then puts the real values back in the model's reply. The post Laranon: Reversible PII Anonymization for Laravel and LLMs appeared first on Laravel News. Join the Laravel Newsletter to get Laravel articles like this directly in your inbox.

## De-identifying text at scale: How Airbnb protects personal data in AI training

DevFeed: [De-identifying text at scale: How Airbnb protects personal data in AI training](<https://devfeed.tech/articles/de-identifying-text-at-scale-how-airbnb-protects-personal-data-in-ai-training-138584.md>)

Original publisher: [Read original article](<https://medium.com/airbnb-engineering/de-identifying-text-at-scale-how-airbnb-protects-personal-data-in-ai-training-32f3ff8ff7d0?source=rss----53c7c27702d5---4>)

Author: Alexandr Climov

Published: 2026-11-17T18:01:03Z

Content type: article

Language: en

Sources: [Airbnb](<https://devfeed.tech/sources/the-airbnb-tech-blog-medium.md>)

Topics: [pii](<https://devfeed.tech/topics/pii.md>), [privacy-filter](<https://devfeed.tech/topics/privacy-filter.md>), [managing large-scale text data in AI](<https://devfeed.tech/topics/managing-large-scale-text-data-in-ai.md>), [Data Quality](<https://devfeed.tech/topics/data-quality.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-training](<https://devfeed.tech/tags/ai-training.md>), [airbnb](<https://devfeed.tech/tags/airbnb.md>), [data](<https://devfeed.tech/tags/data.md>), [de-identification](<https://devfeed.tech/tags/de-identification.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [extensibility](<https://devfeed.tech/tags/extensibility.md>), [hosting](<https://devfeed.tech/tags/hosting.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [messages](<https://devfeed.tech/tags/messages.md>), [modularity](<https://devfeed.tech/tags/modularity.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [personal-data](<https://devfeed.tech/tags/personal-data.md>), [pii](<https://devfeed.tech/tags/pii.md>), [presidio](<https://devfeed.tech/tags/presidio.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [reviews](<https://devfeed.tech/tags/reviews.md>), [support](<https://devfeed.tech/tags/support.md>), [technology](<https://devfeed.tech/tags/technology.md>), [text](<https://devfeed.tech/tags/text.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

Airbnb de-identifies text used for model training with an on-premises Presidio library integrated into its shared ML workflow. The Analyzer detects and labels PII; the Anonymizer resolves overlapping spans and replaces them. Airbnb added phrase-based context matching for company-specific identifiers and uses GLiNER for PERSON and LOCATION because spaCy's broad location tags did not reliably capture full street addresses in its benchmark. Detection thresholds and entity scope vary by consumer. Most text PII receives plausible Faker-generated replacements; per-run secret seeding keeps repeated values consistent within a run while changing fakes across runs, without storing a real-to-fake map. Structured data uses salted hashes for identifiers, truncated coordinates, and format-preserving encryption for internal IDs. Benchmarks report recall, precision, and throughput by entity type and language; releases must avoid regressions. The library's protection depends on pipelines using a current version and compliant configuration, and new languages require labeled data for evaluation.

### Source excerpt

Here's how Airbnb detects and replaces personal data in messages, reviews, and support chats before it reaches a model. By: Emelin Flores, Alex Klimov, Lee Formyduval, Bahador Nooraei Introduction Airbnb is built on trust. Every night that a guest books a stay, or a host welcomes someone into their home, that trust is backed by information they've shared with us, which includes personally identifiable information (PII) such as names, home location, current location, and payment details. We now use transaction information to train machine learning models. Before we do so, we remove PII to minimize the number of copies of this information and prevent leakage into and through our models. As we've written before, Airbnb maintains a shared taxonomy as to what constitutes PII, from commonly used fields such as names and card numbers to less frequently used fields such as precise location coordinates. We use de-identification to remove PII and replace it with simulated data that's as realistic as possible to maximize its usefulness for ML model training. Decisions relating to user privacy start as policy. These policies are set and maintained by Airbnb's Privacy Architecture team, with engagement from Legal, Policy, and Engineering. Engineering then implements policy decisions as code. This post focuses on text de-identification. We'll describe the OSS library we chose to start with, the changes we made to create a functional system for Airbnb, and how we confirm that the system is working. 1. Choosing Presidio as the foundation Many organizations are concerned with protecting PII, so there's a lot of existing best practice and tested code out there that we can use. We chose Presidio, an open-source library originally developed by Microsoft and recently transitioned to community governance. We needed to meet demanding requirements: On-premises hosting, to keep data on our infrastructure General-purpose functionality and high degree of modularity, for maximum extensibility

## How packaged clean rooms bring RampID to your lakehouse

DevFeed: [How packaged clean rooms bring RampID to your lakehouse](<https://devfeed.tech/articles/how-packaged-clean-rooms-bring-rampid-to-your-lakehouse-138351.md>)

Original publisher: [Read original article](<https://www.databricks.com/blog/liveramp-rampid-identity-resolution-packaged-clean-rooms>)

Author: Ryan Werth; Johanna Hughes

Published: 2026-11-17T16:30:00Z

Content type: article

Language: en

Sources: [Databricks](<https://devfeed.tech/sources/databricks.md>)

Topics: [databricks](<https://devfeed.tech/topics/databricks.md>), [pii](<https://devfeed.tech/topics/pii.md>), [identity and access management](<https://devfeed.tech/topics/identity-and-access-management.md>)

Tags: [about](<https://devfeed.tech/tags/about.md>), [activation](<https://devfeed.tech/tags/activation.md>), [added](<https://devfeed.tech/tags/added.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [aliases](<https://devfeed.tech/tags/aliases.md>), [all](<https://devfeed.tech/tags/all.md>), [allows](<https://devfeed.tech/tags/allows.md>), [also](<https://devfeed.tech/tags/also.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [analytics](<https://devfeed.tech/tags/analytics.md>), [api](<https://devfeed.tech/tags/api.md>), [assets](<https://devfeed.tech/tags/assets.md>), [audience-building](<https://devfeed.tech/tags/audience-building.md>), [audit](<https://devfeed.tech/tags/audit.md>), [audit-trail](<https://devfeed.tech/tags/audit-trail.md>), [databricks](<https://devfeed.tech/tags/databricks.md>), [identity](<https://devfeed.tech/tags/identity.md>), [marketing](<https://devfeed.tech/tags/marketing.md>), [partner](<https://devfeed.tech/tags/partner.md>), [partners](<https://devfeed.tech/tags/partners.md>), [pii](<https://devfeed.tech/tags/pii.md>), [platform](<https://devfeed.tech/tags/platform.md>)

### AI overview

LiveRamp Embedded Identity resolves customer records to RampIDs inside a Databricks packaged clean room, avoiding the export-and-match-file round trip. The customer and LiveRamp share Unity Catalog assets read-only through OpenSharing; data stays in each party's storage while provider code and identity graph remain hidden. Unlike an approval-based clean room, the packaged mode has no code-review step, and neither party can see the other's data assets. The customer triggers runs and receives RampID outputs; both parties can see run history and a shared output schema for troubleshooting. Requirements include a LiveRamp agreement and API credentials, a Unity Catalog workspace, CREATE CLEAN ROOM privilege, and prepared PII and metadata tables; SecureConnect is needed for storage behind a firewall or private endpoint. Outputs are retained for 30 days, after which desired results must be copied to customer storage. The clean-room creator pays Jobs Serverless compute for each run.

### Source excerpt

Partner identity resolution used to mean sending your PII to the vendorHistorically on Databricks...

## What can you safely tell AI at work?

DevFeed: [What can you safely tell AI at work?](<https://devfeed.tech/articles/what-can-you-safely-tell-ai-at-work-136230.md>)

Original publisher: [Read original article](<https://www.computerworld.com/article/4244131/what-can-you-safely-tell-ai-at-work.html>)

Author: Pat Brans

Published: 2026-11-16T12:04:12Z

Content type: news

Language: en

Sources: [Computerworld](<https://devfeed.tech/sources/computerworld.md>)

Topics: [shadow AI](<https://devfeed.tech/topics/shadow-ai.md>), [pii](<https://devfeed.tech/topics/pii.md>), [sensitive data](<https://devfeed.tech/topics/sensitive-data.md>), [Access Control](<https://devfeed.tech/topics/access-control.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-at-work](<https://devfeed.tech/tags/ai-at-work.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [artificial-intelligence-generative-ai-it-leadership-it-strategy-laws-and-regulations-privacy-s](<https://devfeed.tech/tags/artificial-intelligence-generative-ai-it-leadership-it-strategy-laws-and-regulations-privacy-s.md>), [code](<https://devfeed.tech/tags/code.md>), [compliance](<https://devfeed.tech/tags/compliance.md>), [data-protection](<https://devfeed.tech/tags/data-protection.md>), [edition-us](<https://devfeed.tech/tags/edition-us.md>), [error-messages](<https://devfeed.tech/tags/error-messages.md>), [generative-ai](<https://devfeed.tech/tags/generative-ai.md>), [governance](<https://devfeed.tech/tags/governance.md>), [it-leadership](<https://devfeed.tech/tags/it-leadership.md>), [it-strategy](<https://devfeed.tech/tags/it-strategy.md>), [language-en](<https://devfeed.tech/tags/language-en.md>), [laws-and-regulations](<https://devfeed.tech/tags/laws-and-regulations.md>), [logs](<https://devfeed.tech/tags/logs.md>), [personally-identifiable-information](<https://devfeed.tech/tags/personally-identifiable-information.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [protection](<https://devfeed.tech/tags/protection.md>), [secrets](<https://devfeed.tech/tags/secrets.md>), [security](<https://devfeed.tech/tags/security.md>), [storytype-feature](<https://devfeed.tech/tags/storytype-feature.md>)

### AI overview

Troubleshooting details can expose sensitive information even after names are removed: service names, file paths, request sequences, customer numbers, meeting context, screenshots, and hidden spreadsheet columns may still identify people or reveal business processes. Sawant recommends reproducing issues with fictional data, simplifying until outsiders could not infer proprietary details, and using internal resources when safe sharing requires too much context. An approved AI tool still has limits; employees need to know what data it may handle and how inputs are stored, used, and accessed. Employers should clarify permitted accounts, tasks, and data through contracts, safeguards, examples, training, access controls, and review. The article also distinguishes confidentiality from policy compliance: using an unapproved tool may violate instructions even without disclosing client information. Asking in writing can document guidance but does not guarantee protection. The excerpt stresses practical support, escalation routes, monitoring, and updating controls as tools gain AI features.

### Source excerpt

When Bhaskar Bharat Sawant troubleshoots a software problem, an error message often provides the first clue. It can also present a dilemma: how much of that message can he safely share with AI tools? Sawant, a lead engineer and solution architect at building-products manufacturer Cornerstone Building Brands, works with legacy applications and newer technologies. GenAI can help explain an error or suggest a way forward. But the input it needs to understand the problem may expose personally identifiable information, a customer's competitive secrets, or his own company's proprietary information. "Logs or error messages can reveal internal service names, configuration details, or business processes," Sawant said. "You have to protect that information whether it belongs to your employer or a customer." A customer's name is an obvious identifier. A method name, file path, or sequence of requests between systems is less conspicuous. Together, those details can expose how an application works or how a business operates -- even when no individual can be identified. Sawant's approach is to construct a smaller, fictional version of the problem. "I create a small code example with made-up data that reproduces the issue," he said. "Then I ask myself whether someone outside the company could piece together proprietary information from it. If they could, I simplify it further or keep it out of the AI tool." That process takes judgment and effort before AI can save him any time. It also exposes the difficulty behind a simple workplace instruction: don't share confidential or sensitive information with generative AI. Employees must translate that warning into decisions about actual work, often while trying to meet a deadline. What remains after you remove the names Employees can recognize the need to protect information and still overlook what they are sharing. Julian Gage, who provides fractional data protection officer services through Engage Compliance, recalls advising a health t

## Amazon Bedrock tokenomics data lake

DevFeed: [Amazon Bedrock tokenomics data lake](<https://devfeed.tech/articles/amazon-bedrock-tokenomics-data-lake-127426.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/machine-learning/amazon-bedrock-tokenomics-data-lake/>)

Author: Steph Gooch

Published: 2026-11-10T16:56:36Z

Content type: article

Language: en

Sources: [Artificial Intelligence](<https://devfeed.tech/sources/artificial-intelligence.md>)

Topics: [LLM observability](<https://devfeed.tech/topics/llm-observability.md>), [data lake](<https://devfeed.tech/topics/data-lake.md>), [streaming data to S3 with Parquet](<https://devfeed.tech/topics/streaming-data-to-s3-with-parquet.md>), [Parquet files in Amazon S3](<https://devfeed.tech/topics/parquet-files-in-amazon-s3.md>), [Amazon CloudWatch](<https://devfeed.tech/topics/amazon-cloudwatch.md>), [Amazon S3](<https://devfeed.tech/topics/amazon-s3.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [AWS Organizations](<https://devfeed.tech/topics/aws-organizations.md>), [Data serialization](<https://devfeed.tech/topics/data-serialization.md>), [elt](<https://devfeed.tech/topics/elt.md>), [pii](<https://devfeed.tech/topics/pii.md>)

Tags: [advanced-300](<https://devfeed.tech/tags/advanced-300.md>), [amazon-athena](<https://devfeed.tech/tags/amazon-athena.md>), [amazon-bedrock](<https://devfeed.tech/tags/amazon-bedrock.md>), [amazon-cloudwatch](<https://devfeed.tech/tags/amazon-cloudwatch.md>), [amazon-cloudwatch-logs](<https://devfeed.tech/tags/amazon-cloudwatch-logs.md>), [amazon-data-firehose](<https://devfeed.tech/tags/amazon-data-firehose.md>), [amazon-s3](<https://devfeed.tech/tags/amazon-s3.md>), [api](<https://devfeed.tech/tags/api.md>), [aws-cloudformation](<https://devfeed.tech/tags/aws-cloudformation.md>), [aws-cost-and-usage-report](<https://devfeed.tech/tags/aws-cost-and-usage-report.md>), [aws-glue](<https://devfeed.tech/tags/aws-glue.md>), [aws-identity-and-access-management-iam](<https://devfeed.tech/tags/aws-identity-and-access-management-iam.md>), [aws-key-management-service-aws-kms](<https://devfeed.tech/tags/aws-key-management-service-aws-kms.md>), [aws-lambda](<https://devfeed.tech/tags/aws-lambda.md>), [aws-quick-sight](<https://devfeed.tech/tags/aws-quick-sight.md>), [cost](<https://devfeed.tech/tags/cost.md>), [data-lake](<https://devfeed.tech/tags/data-lake.md>), [logs](<https://devfeed.tech/tags/logs.md>), [model](<https://devfeed.tech/tags/model.md>), [observability](<https://devfeed.tech/tags/observability.md>), [technical-how-to](<https://devfeed.tech/tags/technical-how-to.md>), [tokenomics](<https://devfeed.tech/tags/tokenomics.md>)

### AI overview

The data lake centralizes Amazon Bedrock invocation logs from linked accounts into an Athena-queryable table, exposing prompts and responses, token counts, model, stop reason, account, timestamp, and IAM identity. CloudWatch subscription filters forward logs to a central account, where Lambda unwraps batches, removes non-invocation messages, and flattens identity data; Data Firehose converts records to Parquet in S3, and Glue defines the Athena table. Deployment uses two CloudFormation stacks: create the central stack first, then source-account stacks, enable Bedrock logging separately in each account, verify data arrival, and run the saved Athena query to create the flattened view. Joining the table to CUR 2.0 by IAM ARN can attribute spend. The S3 bucket uses KMS encryption and IAM policies control access; because prompts may contain PII, permissions need review. The article notes that S3 buckets must be emptied and the Athena workgroup deleted manually during cleanup.

### Source excerpt

Running Amazon Bedrock across many accounts shows you the bill but not what happened inside each call. This post builds a centralized observability data lake that aggregates Bedrock invocation logs from every linked account into one queryable table, so you can see prompts, token counts, model choice, and per-invocation cost across your organization.

## Public, Identifiable, and Effectively Immutable: Ethical Lessons from Research on arXiv

DevFeed: [Public, Identifiable, and Effectively Immutable: Ethical Lessons from Research on arXiv](<https://devfeed.tech/articles/public-identifiable-and-effectively-immutable-ethical-lessons-from-research-on-arxiv-126981.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.12271>)

Author: Jan Pennekamp, Johannes Lohm\\"oller, Peter Mayer, Arjun Arunasalam

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [Cybersecurity](<https://devfeed.tech/topics/cybersecurity.md>), [pii](<https://devfeed.tech/topics/pii.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [attribution](<https://devfeed.tech/tags/attribution.md>), [availability](<https://devfeed.tech/tags/availability.md>), [best-practices](<https://devfeed.tech/tags/best-practices.md>), [central](<https://devfeed.tech/tags/central.md>), [challenge](<https://devfeed.tech/tags/challenge.md>), [challenges](<https://devfeed.tech/tags/challenges.md>), [community](<https://devfeed.tech/tags/community.md>), [confidence](<https://devfeed.tech/tags/confidence.md>), [contributions](<https://devfeed.tech/tags/contributions.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [cybersecurity-research](<https://devfeed.tech/tags/cybersecurity-research.md>), [data](<https://devfeed.tech/tags/data.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [decision](<https://devfeed.tech/tags/decision.md>), [decision-making](<https://devfeed.tech/tags/decision-making.md>), [public](<https://devfeed.tech/tags/public.md>), [research](<https://devfeed.tech/tags/research.md>)

### AI overview

Publicly available research data can still raise ethical concerns when reused for empirical analysis, even when it includes deliberate real-name attribution. The paper examines arXiv source files and related public, live resources, noting risks from personally identifying information, leaked secrets, and other sensitive details that may cause reputational or practical harm. It recommends treating public availability as only the starting point for ethical reasoning, minimizing the amplification of exposed information, calibrating disclosure to a finding's potential harm, confidence, and actionability, and accounting for effective immutability.

### Source excerpt

arXiv:2611.12271v1 Announce Type: new Abstract: Ethical decision-making is central to safe cybersecurity research. Recent studies, one of which is our own, analyzing the source files of arXiv papers bring this challenge to light - arXiv publicly distributes the source files, to the surprise of many authors. Unique aspects reveal ethical challenges and dilemmas that extend beyond established ethical frameworks, institutional processes, and community best practices. In this work, we examine this case and provide insights for similar research on other public live resources, such as Zenodo or the Wayback Machine. While such empirical studies are particularly realistic, they must account for personally identifiable information, leaked secrets, and other sensitive details that could expose, shame, or otherwise negatively affect individuals, possibly leading to both reputational and practical harm. sDrawing on our collective experience and discussions among researchers, affected authors, and readers, we make three contributions: First, we define the ethical setting and highlight that even intentionally-public data with deliberate real-name attribution can raise ethical concerns when repurposed for empirical analysis. Second, we discuss the ethical issues surrounding the use of arXiv data and derive four higher-level dilemmas that researchers may face when working with similar data. Third, we derive practical guidance for ethically navigating such settings and, in doing so, distill four lessons for research on arXiv and similar public, live datasets: public availability should serve as the starting point of ethical reasoning; research should minimize the amplification it induces on exposed information; disclosure should be calibrated to the harm, confidence, and actionability of a finding; and effective immutability must be accounted for. Together, these contribute to community resources for ethically-sound cybersecurity experiments.

## Redacting Personal Data: A Speed and Accuracy Comparison of Regex, Presidio and Hybrid Designs

DevFeed: [Redacting Personal Data: A Speed and Accuracy Comparison of Regex, Presidio and Hybrid Designs](<https://devfeed.tech/articles/redacting-personal-data-a-speed-and-accuracy-comparison-of-regex-presidio-and-hybrid-designs-120219.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.07633>)

Author: Sai Yashwant, Manoj Kumar Gondimalla, Arjun Sharma, Shubham

Published: 2026-11-06T05:00:00Z

Content type: comparison

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [pii](<https://devfeed.tech/topics/pii.md>), [ner](<https://devfeed.tech/topics/ner.md>), [ai security](<https://devfeed.tech/topics/ai-security.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>)

Tags: [comparison](<https://devfeed.tech/tags/comparison.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [data](<https://devfeed.tech/tags/data.md>), [fast](<https://devfeed.tech/tags/fast.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [latency](<https://devfeed.tech/tags/latency.md>), [personal](<https://devfeed.tech/tags/personal.md>), [personal-data](<https://devfeed.tech/tags/personal-data.md>), [pii](<https://devfeed.tech/tags/pii.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [redaction](<https://devfeed.tech/tags/redaction.md>), [regex](<https://devfeed.tech/tags/regex.md>), [resume](<https://devfeed.tech/tags/resume.md>), [spacy](<https://devfeed.tech/tags/spacy.md>), [speed](<https://devfeed.tech/tags/speed.md>)

### AI overview

In tests on 100 synthetic resumes, regex redaction was fastest (0.20 ms median) but detected only 88.9% of PII, missing names not known in advance. Microsoft Presidio with spaCy, a hybrid using Presidio for names and regex for structured identifiers, and a fast hybrid calling spaCy's named-entity component each detected all PII and preserved all business terms. Their median latencies were 37.59 ms, 30.23 ms, and 13.48 ms, respectively; even Presidio added just 0.41% against the workflow's 9,120 ms median LLM call. All designs also removed the known subject and used consistent placeholders. The paper recommends the fast hybrid as a default for personal-data pipelines, with regex as a safety net. Results are from this synthetic resume workflow.

### Source excerpt

arXiv:2611.07633v1 Announce Type: new Abstract: Many enterprise projects now pass free text to large language models (LLMs), and that text often contains personally identifiable information (PII) such as names, phone numbers, email addresses, home addresses and dates of birth. Once a prompt is sent, this data is copied into provider systems, application logs and observability traces. Redacting PII before the call is an obvious control, yet teams seldom have numbers on what it costs in latency or what it lets through. This paper compares four redaction designs that any LLM pipeline can adopt: a pure regular-expression (regex) redactor, Microsoft Presidio running on a spaCy model, a hybrid in which Presidio finds only person names while regex handles structured identifiers, and a fast hybrid that calls spaCy's named-entity component directly. All four share a known-entity rule, which removes the person the system already knows it is handling, and a consistent placeholder scheme, so the LLM can still tell one person from another after redaction. We evaluate the designs on a sample workflow, which we call ProfileAssist, that turns resumes into structured profiles, using 100 synthetic resumes (523 characters on average, with 900 PII values and 600 business terms that must survive). Regex ran at a 0.20 ms median but caught only 88.9% of PII, because it cannot find names that are not known in advance, such as the references listed in a resume. Presidio, hybrid and fast hybrid each caught 100% of PII and kept 100% of business terms, at median latencies of 37.59 ms, 30.23 ms and 13.48 ms. Measured against a 9,120 ms median LLM call, even the slowest option adds only 0.41% to the request. We therefore recommend the fast hybrid as a default for LLM pipelines that handle personal data, with regex kept as a safety net rather than as the only control.

## DigiCert data breach exposes customer information, identity documents

DevFeed: [DigiCert data breach exposes customer information, identity documents](<https://devfeed.tech/articles/digicert-data-breach-exposes-customer-information-identity-documents-119222.md>)

Original publisher: [Read original article](<https://www.bleepingcomputer.com/news/security/digicert-data-breach-exposes-customer-information-identity-documents/>)

Author: Lawrence Abrams

Published: 2026-11-05T16:12:46Z

Content type: news

Language: en

Sources: [BleepingComputer](<https://devfeed.tech/sources/bleepingcomputer.md>)

Topics: [enterprise data protection](<https://devfeed.tech/topics/enterprise-data-protection.md>), [pii](<https://devfeed.tech/topics/pii.md>), [Authorization](<https://devfeed.tech/topics/authorization.md>), [Access Control](<https://devfeed.tech/topics/access-control.md>), [identity and access management](<https://devfeed.tech/topics/identity-and-access-management.md>), [Cybersecurity](<https://devfeed.tech/topics/cybersecurity.md>)

Tags: [breach](<https://devfeed.tech/tags/breach.md>), [computer-help](<https://devfeed.tech/tags/computer-help.md>), [computer-security](<https://devfeed.tech/tags/computer-security.md>), [computers](<https://devfeed.tech/tags/computers.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-data](<https://devfeed.tech/tags/customer-data.md>), [data](<https://devfeed.tech/tags/data.md>), [data-breach](<https://devfeed.tech/tags/data-breach.md>), [data-leak](<https://devfeed.tech/tags/data-leak.md>), [digicert](<https://devfeed.tech/tags/digicert.md>), [identity-documents](<https://devfeed.tech/tags/identity-documents.md>), [infosec](<https://devfeed.tech/tags/infosec.md>), [infosec-computer-security](<https://devfeed.tech/tags/infosec-computer-security.md>), [linux](<https://devfeed.tech/tags/linux.md>), [mac](<https://devfeed.tech/tags/mac.md>), [malware](<https://devfeed.tech/tags/malware.md>), [malware-removal](<https://devfeed.tech/tags/malware-removal.md>), [mfa](<https://devfeed.tech/tags/mfa.md>), [passport](<https://devfeed.tech/tags/passport.md>), [security](<https://devfeed.tech/tags/security.md>), [spyware](<https://devfeed.tech/tags/spyware.md>), [support](<https://devfeed.tech/tags/support.md>), [tech-support](<https://devfeed.tech/tags/tech-support.md>), [technical-support](<https://devfeed.tech/tags/technical-support.md>), [virus](<https://devfeed.tech/tags/virus.md>), [virus-removal](<https://devfeed.tech/tags/virus-removal.md>), [windows](<https://devfeed.tech/tags/windows.md>)

### AI overview

A February 2025 server migration exposed a DigiCert internal business system that had previously been VPN-only. Its self-registration feature and an authorization flaw let external accounts read archived customer data. DigiCert found unauthorized accounts in June 2026 but confirmed access to customer information only after a further investigation on October 19; the account accessed records between June 27 and July 2. The affected data included names, business contact details, addresses, and, for some people, identity documents such as passports or driver's licenses. DigiCert says access was read-only and could not alter records, validate or issue certificates. It restricted system access, disabled self-registration, and added controls. The company says its certificate services, CertCentral, DigiCert ONE, customer credentials, and issued certificates were not compromised. The excerpt does not state how many people were affected.

### Source excerpt

DigiCert warns it suffered a data breach after a server migration mistakenly exposed an internal business system to the internet, allowing unauthorized access to customer information and identity documents. [...]

## CVE-2026-93389 - LiteSpeed Cache \<= 7.9.1 - Unauthenticated Sensitive Information Disclosure (WooCommerce Store API Cart/Checkout PII + Cart-Token) via Public REST Caching

DevFeed: [CVE-2026-93389 - LiteSpeed Cache \<= 7.9.1 - Unauthenticated Sensitive Information Disclosure (WooCommerce Store API Cart/Checkout PII + Cart-Token) via Public REST Caching](<https://devfeed.tech/articles/cve-2026-93389-litespeed-cache-7-9-1-unauthenticated-sensitive-information-disclosure-woocommerce-store-api-cart-checkout-pii-cart-token-via-public-rest-caching-118154.md>)

Original publisher: [Read original article](<https://cvefeed.io/vuln/detail/CVE-2026-93389>)

Author: CVEFEED.io; wordfence.com

Published: 2026-11-05T05:18:40Z

Content type: news

Language: en

Sources: [Latest High/Critical Vulnerabilitiy Feed](<https://devfeed.tech/sources/latest-high-critical-vulnerabilitiy-feed.md>)

Topics: [Cache](<https://devfeed.tech/topics/cache.md>), [pii](<https://devfeed.tech/topics/pii.md>), [REST API](<https://devfeed.tech/topics/rest-api.md>), [JSON Web Tokens](<https://devfeed.tech/topics/jwt.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [api](<https://devfeed.tech/tags/api.md>), [billing](<https://devfeed.tech/tags/billing.md>), [cache](<https://devfeed.tech/tags/cache.md>), [caching](<https://devfeed.tech/tags/caching.md>), [checkout](<https://devfeed.tech/tags/checkout.md>), [cve](<https://devfeed.tech/tags/cve.md>), [cve-2026-93389](<https://devfeed.tech/tags/cve-2026-93389.md>), [disclosure](<https://devfeed.tech/tags/disclosure.md>), [information](<https://devfeed.tech/tags/information.md>), [litespeed-cache](<https://devfeed.tech/tags/litespeed-cache.md>), [litespeedtech](<https://devfeed.tech/tags/litespeedtech.md>), [pii](<https://devfeed.tech/tags/pii.md>), [public](<https://devfeed.tech/tags/public.md>), [rest](<https://devfeed.tech/tags/rest.md>), [store](<https://devfeed.tech/tags/store.md>), [token](<https://devfeed.tech/tags/token.md>), [woocommerce](<https://devfeed.tech/tags/woocommerce.md>)

### AI overview

LiteSpeed Cache versions up to and including 7.9.1 can expose WooCommerce customer data when the shipped default `cache-rest=1` setting is enabled. The excerpt says Store API cart and checkout GET responses are cached publicly without varying by Cart-Token or WooCommerce session, so a later unauthenticated client may receive a shopper's personalized response. Disclosed data may include billing and shipping details, cart contents and totals, a live Cart-Token JWT and nonce, and checkout `order_key`; the source says the token and nonce can enable cart or checkout session hijacking. The excerpt gives a CVSS severity of 8.2 (HIGH) but provides no remediation steps.

### Source excerpt

CVE ID :CVE-2026-93389 Published : Nov. 5, 2026, 5:18 a.m. | 2 hours, 9 minutes ago Description :The LiteSpeed Cache plugin for WordPress is vulnerable to unauthenticated disclosure of sensitive WooCommerce customer data in all versions up to, and including, 7.9.1 With Cache REST enabled (the shipped default cache-rest=1), LiteSpeed stores WooCommerce Store API GET responses (/wc/store/v1/cart and /wc/store/v1/checkout) as PUBLIC cache objects and does not vary the cache on the Cart-Token or WooCommerce session. A shopper's personalized cart/checkout response is therefore served to any later, cookie-less and token-less client as an LSCache HIT. This makes it possible for unauthenticated attackers to read another shopper's billing and shipping details (name, company, street address, email, phone), cart contents and totals, and the shopper's live Cart-Token session JWT and nonce (enabling cart/checkout-session hijack), plus order_key on the checkout endpoint. Severity: 8.2 | HIGH Visit the link for more details, such as CVSS details, affected products, timeline, and more...

## AI-Linked Breaches Mount as WorkOS, Northflank, Trigger.dev Disclose Hacks

DevFeed: [AI-Linked Breaches Mount as WorkOS, Northflank, Trigger.dev Disclose Hacks](<https://devfeed.tech/articles/ai-linked-breaches-mount-as-workos-northflank-trigger-dev-disclose-hacks-116261.md>)

Original publisher: [Read original article](<https://www.securityweek.com/ai-linked-breaches-mount-as-workos-northflank-trigger-dev-disclose-hacks/>)

Author: Ionut Arghire

Published: 2026-11-04T12:43:20Z

Content type: news

Language: en

Sources: [SecurityWeek](<https://devfeed.tech/sources/securityweek.md>)

Topics: [ai-assisted attacks](<https://devfeed.tech/topics/ai-assisted-attacks.md>), [Cybersecurity](<https://devfeed.tech/topics/cybersecurity.md>), [Credential theft](<https://devfeed.tech/topics/credential-theft.md>), [Secrets Management](<https://devfeed.tech/topics/secrets-management.md>), [pii](<https://devfeed.tech/topics/pii.md>), [Vulnerabilities](<https://devfeed.tech/topics/vulnerabilities.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>)

Tags: [access-tokens](<https://devfeed.tech/tags/access-tokens.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-assisted](<https://devfeed.tech/tags/ai-assisted.md>), [ai-tools](<https://devfeed.tech/tags/ai-tools.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [attacks](<https://devfeed.tech/tags/attacks.md>), [automate](<https://devfeed.tech/tags/automate.md>), [breaches](<https://devfeed.tech/tags/breaches.md>), [featured](<https://devfeed.tech/tags/featured.md>), [hackers](<https://devfeed.tech/tags/hackers.md>), [hacks](<https://devfeed.tech/tags/hacks.md>), [malware-threats](<https://devfeed.tech/tags/malware-threats.md>), [northflank](<https://devfeed.tech/tags/northflank.md>), [offensive](<https://devfeed.tech/tags/offensive.md>), [offensive-tools](<https://devfeed.tech/tags/offensive-tools.md>), [tokens](<https://devfeed.tech/tags/tokens.md>), [tools](<https://devfeed.tech/tags/tools.md>), [trigger-dev](<https://devfeed.tech/tags/trigger-dev.md>), [workos](<https://devfeed.tech/tags/workos.md>)

### AI overview

Recent attacks on AI developer platforms appear to share a motive--stealing stored AI credentials for resale or abuse--but the excerpt says the incidents are not linked and initial access varied, including compromised credentials, third-party components, and exploited vulnerabilities. WorkOS and Chainstack disclosed attacks involving the open-source AI penetration-testing framework Artex. Chainstack said attackers entered through a third-party component and accessed customer account data, password hashes, and systems storing node credentials; it urged customers to rotate node credentials. WorkOS said attackers compromised an internal sales application and accessed business contact details, account histories, sales and customer relationship information, communications, and product-use information. Wiz attributed rising demand for stolen keys to LLM-jacking, broker infrastructure, and model distillation, and reported multiple threat actors using some of the same compromised keys.

### Source excerpt

Hackers have been harvesting AI access tokens and have been using AI-assisted offensive tools to automate attacks. The post AI-Linked Breaches Mount as WorkOS, Northflank, Trigger.dev Disclose Hacks appeared first on SecurityWeek.

## Reducing Logprob Without Causal Verification: Diagnosing SAE-Based PII Suppression in LLMs

DevFeed: [Reducing Logprob Without Causal Verification: Diagnosing SAE-Based PII Suppression in LLMs](<https://devfeed.tech/articles/reducing-logprob-without-causal-verification-diagnosing-sae-based-pii-suppression-in-llms-117524.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.04782>)

Author: Chaerin Lee, Wonho Lee, Sohee Park, Daeseon Choi

Published: 2026-11-04T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [pii](<https://devfeed.tech/topics/pii.md>), [causality](<https://devfeed.tech/topics/causality.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>)

Tags: [activation](<https://devfeed.tech/tags/activation.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [circuit](<https://devfeed.tech/tags/circuit.md>), [compare](<https://devfeed.tech/tags/compare.md>), [correlation](<https://devfeed.tech/tags/correlation.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [effect](<https://devfeed.tech/tags/effect.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [evaluations](<https://devfeed.tech/tags/evaluations.md>), [failure](<https://devfeed.tech/tags/failure.md>), [feature](<https://devfeed.tech/tags/feature.md>), [feature-selection](<https://devfeed.tech/tags/feature-selection.md>), [features](<https://devfeed.tech/tags/features.md>), [framework](<https://devfeed.tech/tags/framework.md>), [layer](<https://devfeed.tech/tags/layer.md>), [llms](<https://devfeed.tech/tags/llms.md>), [localization](<https://devfeed.tech/tags/localization.md>), [magnitude](<https://devfeed.tech/tags/magnitude.md>), [pii](<https://devfeed.tech/tags/pii.md>), [verification](<https://devfeed.tech/tags/verification.md>)

### AI overview

Log-probability reduction alone does not establish that SAE interventions specifically suppress PII leakage: large activation perturbations can produce a direction-agnostic reduction. In 36 model, layer, and PII-type cases, 17 interventions appeared successful by log-probability alone, but only 1 passed the authors' multimetric evaluation against coherence-matched random directions. The reported failures included random directions outperforming selected SAE features and feature selection collapsing across PII types. A 9,000-run causal mediation feasibility study found that its verifier separated candidates reliably, while correlation-selected candidates rarely passed. The authors argue for circuit-level localization and causal mediation to evaluate causally grounded PII suppression.

### Source excerpt

arXiv:2611.04782v1 Announce Type: new Abstract: SAE-based feature-level interventions to suppress PII leakage in LLMs are primarily evaluated by log-probability reduction. We show that this criterion cannot distinguish feature specific suppression from a magnitude artifact: a direction-agnostic effect of sufficiently large activation perturbations. Across 36 (model, layer, PII type) cases, we compare selected SAE feature interventions with coherencematched random direction controls using logprob and rank-aware metrics. While 17 cases (47%) appear successful under logprob-only evaluation, only 1 case (2.8%) passes our multimetric framework. The rest reveal two failure modes: random directions outperform selected features, and feature selection collapse across PII types. These results indicate that existing evaluations cannot verify whether interventions target causal pathways of PII leakage. A 9,000 run causal mediation feasibility study corroborates this: the verifier separates candidates reliably, but correlation-selected candidates rarely pass it. We therefore propose a new research mission for NLP: causally grounded PII suppression through circuit-level localization and causal mediation analysis.

## 42AI comes out of stealth with $62.5M and a world model that can 'predict human behavior'

DevFeed: [42AI comes out of stealth with $62.5M and a world model that can 'predict human behavior'](<https://devfeed.tech/articles/42ai-comes-out-of-stealth-with-62-5m-and-a-world-model-that-can-predict-human-behavior-112480.md>)

Original publisher: [Read original article](<https://techcrunch.com/2026/11/02/42ai-comes-out-of-stealth-with-62-5m-and-a-world-model-that-can-predict-human-behavior/>)

Author: Ram Iyer

Published: 2026-11-02T16:30:00Z

Content type: news

Language: en

Sources: [TechCrunch | AI News & Artificial Intelligence](<https://devfeed.tech/sources/techcrunch-ai-news-artificial-intelligence.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [World models](<https://devfeed.tech/topics/world-models.md>), [pii](<https://devfeed.tech/topics/pii.md>), [AI in cybersecurity](<https://devfeed.tech/topics/ai-in-cybersecurity.md>)

Tags: [42ai](<https://devfeed.tech/tags/42ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [behavior](<https://devfeed.tech/tags/behavior.md>), [behavioral-analysis](<https://devfeed.tech/tags/behavioral-analysis.md>), [behavioral-world-model](<https://devfeed.tech/tags/behavioral-world-model.md>), [brokerages](<https://devfeed.tech/tags/brokerages.md>), [chemistry](<https://devfeed.tech/tags/chemistry.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [data](<https://devfeed.tech/tags/data.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [deployments](<https://devfeed.tech/tags/deployments.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [domain](<https://devfeed.tech/tags/domain.md>), [ehud-ezra](<https://devfeed.tech/tags/ehud-ezra.md>), [ethics](<https://devfeed.tech/tags/ethics.md>), [exclusive](<https://devfeed.tech/tags/exclusive.md>), [exploring](<https://devfeed.tech/tags/exploring.md>), [finance](<https://devfeed.tech/tags/finance.md>), [financial](<https://devfeed.tech/tags/financial.md>), [model](<https://devfeed.tech/tags/model.md>), [predictive-analytics](<https://devfeed.tech/tags/predictive-analytics.md>), [startups](<https://devfeed.tech/tags/startups.md>), [world-model](<https://devfeed.tech/tags/world-model.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

42AI says its behavioral world model estimates probabilities of individual traders' decisions by turning signals from onboarding, support interactions, transactions and market activity into trait representations, then simulating choices while preserving underlying tendencies. The company reports use at brokerages for liquidity, margin and hedging, with trading-direction and volume F1 scores of 0.8-0.85. Its claimed 98% accuracy describes whether trading flows fall within a predicted range, not correct prediction of each trade. The model can miss unobserved context, such as advice from friends, and needs domain-specific training for uses beyond finance. Although 42AI says it does not need personally identifiable information, it builds behavioral profiles from linked customer activity; the article leaves open how customers are prevented from using predictions to their own advantage.

### Source excerpt

Here's a question straight out of sci-fi: What if AI could predict what you were going to do before you decided to, even in a completely new situation?

## A gift to scammers: why you should probably keep your birthday a secret

DevFeed: [A gift to scammers: why you should probably keep your birthday a secret](<https://devfeed.tech/articles/a-gift-to-scammers-why-you-should-probably-keep-your-birthday-a-secret-110773.md>)

Original publisher: [Read original article](<https://www.theguardian.com/money/2026/nov/01/do-not-share-date-of-birth-online-scam-fraudsters-advice>)

Author: Shane Hickey

Published: 2026-11-01T07:00:05Z

Content type: article

Language: en

Sources: [The Guardian | Technology](<https://devfeed.tech/sources/the-guardian-technology.md>)

Topics: [online privacy](<https://devfeed.tech/topics/online-privacy.md>), [pii](<https://devfeed.tech/topics/pii.md>), [Social engineering](<https://devfeed.tech/topics/social-engineering.md>), [Cybercrime](<https://devfeed.tech/topics/cybercrime.md>)

Tags: [account](<https://devfeed.tech/tags/account.md>), [bank](<https://devfeed.tech/tags/bank.md>), [birthday](<https://devfeed.tech/tags/birthday.md>), [consumer-affairs](<https://devfeed.tech/tags/consumer-affairs.md>), [cybercrime](<https://devfeed.tech/tags/cybercrime.md>), [data-and-computer-security](<https://devfeed.tech/tags/data-and-computer-security.md>), [e-commerce](<https://devfeed.tech/tags/e-commerce.md>), [fake](<https://devfeed.tech/tags/fake.md>), [fraud](<https://devfeed.tech/tags/fraud.md>), [hacking](<https://devfeed.tech/tags/hacking.md>), [identity](<https://devfeed.tech/tags/identity.md>), [identity-fraud](<https://devfeed.tech/tags/identity-fraud.md>), [identity-theft](<https://devfeed.tech/tags/identity-theft.md>), [information](<https://devfeed.tech/tags/information.md>), [leaving](<https://devfeed.tech/tags/leaving.md>), [money](<https://devfeed.tech/tags/money.md>), [money-identity-fraud-scams-consumer-affairs-cybercrime-hacking-online-shopping-technology-e-commerce](<https://devfeed.tech/tags/money-identity-fraud-scams-consumer-affairs-cybercrime-hacking-online-shopping-technology-e-commerce.md>), [online](<https://devfeed.tech/tags/online.md>), [online-shopping](<https://devfeed.tech/tags/online-shopping.md>), [retail-industry](<https://devfeed.tech/tags/retail-industry.md>), [scams](<https://devfeed.tech/tags/scams.md>), [secret](<https://devfeed.tech/tags/secret.md>), [technology](<https://devfeed.tech/tags/technology.md>)

### AI overview

A date of birth can help scammers build a convincing identity profile when combined with names, contact details and other leaked or public information. The article advises leaving birthday fields blank on low-risk retail and other nonessential accounts; if a site requires one, using a consistent fake date, and considering a separate shopping email address. Keep your real name and DOB for services that verify identity against official records, such as banks, lenders and government agencies. Be wary of unexpected callers or messages even when they know personal details. The article notes that age-restricted purchases may require age verification.

### Source excerpt

Using a fake DOB or leaving the field blank when making an online account can offer protection from identity theft You get a call from your bank saying there has been suspicious activity on your account. They read out your name, address and date of birth so you are happy it is a legitimate caller. But you are talking to criminals who have used your date of birth as part of a web of verifying information to convince you they are someone else. Continue reading...

## How to stop PII and national IDs from reaching LLMs

DevFeed: [How to stop PII and national IDs from reaching LLMs](<https://devfeed.tech/articles/how-to-stop-pii-and-national-ids-from-reaching-llms-119450.md>)

Original publisher: [Read original article](<https://nirmata.com/2026/10/29/how-to-stop-pii-and-national-ids-from-reaching-llms/>)

Author: Ritesh Patel

Published: 2026-10-29T19:00:00Z

Content type: tutorial

Language: en

Sources: [Nirmata](<https://devfeed.tech/sources/nirmata.md>)

Topics: [pii](<https://devfeed.tech/topics/pii.md>), [Claude Code](<https://devfeed.tech/topics/claude-code.md>), [ner](<https://devfeed.tech/topics/ner.md>), [gateway](<https://devfeed.tech/topics/gateway.md>), [Regular expression](<https://devfeed.tech/topics/regular-expression.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-governance](<https://devfeed.tech/tags/ai-governance.md>), [block](<https://devfeed.tech/tags/block.md>), [checksums](<https://devfeed.tech/tags/checksums.md>), [claude-code](<https://devfeed.tech/tags/claude-code.md>), [claude-code-hooks](<https://devfeed.tech/tags/claude-code-hooks.md>), [data-protection](<https://devfeed.tech/tags/data-protection.md>), [gateway](<https://devfeed.tech/tags/gateway.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [ids](<https://devfeed.tech/tags/ids.md>), [llms](<https://devfeed.tech/tags/llms.md>), [ner](<https://devfeed.tech/tags/ner.md>), [pii](<https://devfeed.tech/tags/pii.md>), [security](<https://devfeed.tech/tags/security.md>), [stop](<https://devfeed.tech/tags/stop.md>)

### AI overview

Keep PII from reaching an LLM by checking prompts and tool results before they leave your network, using checksum recognizers for structured IDs and named-entity models for names, then redacting, holding or blocking matches. Coverage depends on checking each route: Claude Code's UserPromptSubmit can block but cannot rewrite typed prompts; PreToolUse can alter tool inputs, and PostToolUse can redact results only after the tool has run. A gateway can inspect model requests routed through it, but separate scripts or MCP servers may bypass it, and gateway redaction support must be verified. Presidio's Aadhaar recognizer is disabled by default and must be added explicitly; its checksum still allows about 10% of random 12-digit numbers through, so matches need review. Detection is not complete, and hook timeouts pass prompts unless onFailure is set to block. The excerpt recommends testing with synthetic valid and near-miss numbers, starting with redaction and review, and limiting agent file access.

### Source excerpt

Stop PII and national IDs reaching LLMs: detect them with checksums and NER, then redact, hold or block each match in Claude Code hooks or at a gateway.

## How to redact PII and PHI before it reaches an LLM

DevFeed: [How to redact PII and PHI before it reaches an LLM](<https://devfeed.tech/articles/how-to-redact-pii-and-phi-before-it-reaches-an-llm-103577.md>)

Original publisher: [Read original article](<https://www.lyzr.ai/blog/redact-pii-and-phi-before-it-reaches-an-llm/>)

Author: Lyzr Team

Published: 2026-10-28T07:44:04Z

Content type: tutorial

Language: en

Sources: [Lyzr.ai](<https://devfeed.tech/sources/lyzr-ai.md>)

Topics: [pii](<https://devfeed.tech/topics/pii.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [llm](<https://devfeed.tech/tags/llm.md>), [pii](<https://devfeed.tech/tags/pii.md>)

### AI overview

Redacting prompts alone does not protect PII and PHI: identifiers can also leak through model responses, tool calls, streams, logs, or agents that bypass the checks. The excerpt recommends layered detection--validated regex for structured values, named entity recognition for names and places, and custom health-record rules for identifiers such as MRNs and care-related dates--followed by typed placeholders. If values must be restored, use random session-scoped tokens with mappings kept in an encrypted vault; hashes may remain personal data, and HIPAA bars re-identification codes derived from the person's information. Run detection within the trust boundary or under the same BAA and region, and apply checks to inputs and outputs through a single gateway. Test misses by identifier type with labeled examples, begin new rules in log-only mode, and rerun evaluations when detectors, prompts, or models change. The excerpt notes that NER can miss sensitive data or misclassify entities, so review and tuning remain necessary.

### Source excerpt

TL;DR Your discharge-summary agent is live. A nurse pastes in a note: "Pt in 412, MRN 0048291, daughter Priya called re: d/c on 3/14, cb...

## ReCoPII: A Relation-Grounded Framework for Building Contextual PII Masking Datasets

DevFeed: [ReCoPII: A Relation-Grounded Framework for Building Contextual PII Masking Datasets](<https://devfeed.tech/articles/recopii-a-relation-grounded-framework-for-building-contextual-pii-masking-datasets-103057.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.35925>)

Author: Taehwan Lee

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [privacy-filter](<https://devfeed.tech/topics/privacy-filter.md>), [pii](<https://devfeed.tech/topics/pii.md>), [responsible-ai](<https://devfeed.tech/topics/responsible-ai.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [framework](<https://devfeed.tech/tags/framework.md>), [masking](<https://devfeed.tech/tags/masking.md>), [pii](<https://devfeed.tech/tags/pii.md>)

### AI overview

ReCoPII builds contextual PII masking data without assuming a document has only one data subject. It plans ontology-constrained entity relations, creates placeholder-based insertion sentences, reconstructs annotations with code, checks them with a separate LLM, and assigns MASK/KEEP labels from the relations. On Korean documents across six domains, the resulting ReCoPII-Ko corpus achieved 88.42% annotator agreement on sampled KEEP labels across 300 pairs, compared with 26.27% for SPY-style augmentation from the same sources. Models trained on ReCoPII-Ko over-masked less on a matched test. Jointly fine-tuning contextual masking with a PII detector reduced most over-masking from detection alone at a fixed leakage budget; among four detectors meeting that budget, over-masking was at most 5.04% and leakage at most 4.18% on human-verified entity records. The detector was comparable to XLM-R-large on KDPII with about one quarter as many parameters.

### Source excerpt

arXiv:2610.35925v1 Announce Type: cross Abstract: Training contextual masking systems for personally identifiable information (PII) requires examples distinguishing maskable from preservable mentions of one entity type, but privacy restrictions limit access to real data. Author-based augmentation (SPY) labels mentions unrelated to the author as non-PII, presuming a single data subject and overlooking others. We propose ReCoPII (Relation-grounded Contextual PII masking), a framework that builds such data without this premise: it plans ontology-constrained entity relations, writes insertion sentences with placeholders, reconstructs annotations with code, checks the result with a separate LLM, and derives MASK/KEEP labels from entity relations. On Korean documents from six domains, it yields ReCoPII-Ko, a corpus with relation, privacy, and MASK/KEEP annotations, which we validate. Across 300 pairs, four annotators agree with 88.42% of its sampled KEEP labels, against 26.27% for SPY-style augmentation of the same sources. Classifiers trained on it over-mask less than under that reference on a matched test. Contextual masking fine-tuned jointly with a PII detector removes most over-masking of detection alone within a fixed leakage budget; on human-verified entity records, the four detectors meeting it over-mask at most 5.04% with at most 4.18% leakage. The detector is comparable to XLM-R-large on KDPII with about a quarter of its parameters.

## PostgreSQL Anonymizer 3.3 : Major Security Update

DevFeed: [PostgreSQL Anonymizer 3.3 : Major Security Update](<https://devfeed.tech/articles/postgresql-anonymizer-3-3-major-security-update-105461.md>)

Original publisher: [Read original article](<https://www.postgresql.org/about/news/postgresql-anonymizer-33-major-security-update-3423/>)

Published: 2026-10-28T00:00:00Z

Content type: release

Language: en

Sources: [PostgreSQL news](<https://devfeed.tech/sources/postgresql-news.md>)

Topics: [Security](<https://devfeed.tech/topics/security.md>), [pii](<https://devfeed.tech/topics/pii.md>), [sensitive data](<https://devfeed.tech/topics/sensitive-data.md>), [vulnerability management](<https://devfeed.tech/topics/vulnerability-management.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [critical](<https://devfeed.tech/tags/critical.md>), [data](<https://devfeed.tech/tags/data.md>), [database](<https://devfeed.tech/tags/database.md>), [dbaas](<https://devfeed.tech/tags/dbaas.md>), [debian](<https://devfeed.tech/tags/debian.md>), [docker](<https://devfeed.tech/tags/docker.md>), [extension](<https://devfeed.tech/tags/extension.md>), [masking](<https://devfeed.tech/tags/masking.md>), [personally-identifiable-information](<https://devfeed.tech/tags/personally-identifiable-information.md>), [pii](<https://devfeed.tech/tags/pii.md>), [postgres](<https://devfeed.tech/tags/postgres.md>), [postgresql](<https://devfeed.tech/tags/postgresql.md>), [security](<https://devfeed.tech/tags/security.md>), [security-fixes](<https://devfeed.tech/tags/security-fixes.md>), [security-update](<https://devfeed.tech/tags/security-update.md>), [sensitive-data](<https://devfeed.tech/tags/sensitive-data.md>), [sql](<https://devfeed.tech/tags/sql.md>), [strategies](<https://devfeed.tech/tags/strategies.md>), [update](<https://devfeed.tech/tags/update.md>)

### AI overview

PostgreSQL Anonymizer 3.3 fixes 12 critical vulnerabilities that could, in some circumstances, let users gain superuser privileges or masked users access personal information. The release adds restrictions against malicious behavior. The source urges all users to upgrade to 3.3 as soon as possible and strongly advises changing the `anon.salt` GUC parameter; it points to the UPGRADE section for details. The extension masks personally identifiable or commercially sensitive data using six strategies and functions such as substitution, pseudonymization, scrambling, and noise addition. The excerpt names package, Ansible, and Docker installation options and several DBaaS providers and PostgreSQL forks, but does not include the full vulnerability list or detailed upgrade instructions.

### Source excerpt

Eymoutiers, France, October 27th, 2026 Dalibo is pleased to announce PostgreSQL Anonymizer 3.3 with a panel of important security fixes. Enhanced Privacy Protection for Your Data PostgreSQL Anonymizer is an extension that hides or replaces personally identifiable information (PII) or commercially sensitive data from a PostgreSQL database. The extension offers 6 different masking strategies: Dynamic Masking - Real-time data protection Static Masking - Permanent data transformation Replica Masking - Anonymized logical replication Backup Masking - Privacy-protected database exports Masking Views - Controlled data visibility Masking Data Wrappers - Extended protection across systems Each strategy is complemented by an enhanced suite of Masking Functions, including advanced techniques such as: Substitution, Randomization, Faking, Pseudonymization, Partial Scrambling, Shuffling, Noise Addition and Generalization. The extension can be installed with Debian and RPM packages, an Ansible role, a Docker image, etc. You can use it on most major DBaaS providers including : Alibaba Cloud, Crunchy Bridge, Google Cloud SQL, IBM Cloud, Microsoft Azure Database, Neon, Yandex It is also available on some Postgres forks such as EDB Advanced Postgres, Greenplum and Yugabyte. See the INSTALL section of the documentation for more details! Important Security Update Version 3.3 includes fixes for 12 critical vulnerabilities allowing some users to gain superuser privileges under certain circumstances or allowing masked users to access personal information. These vulnerabilities are declared below: CVE-2026-110181: Masked roles read the authentic data through SQL functions with a SQL-standard body CVE-2026-105615: Masked roles can read pg_stats CVE-2026-105616: The secret salt is readable in pg_db_role_setting CVE-2026-110183: After the static masking, pg_stats keeps the authentic values of the masked columns CVE-2026-105619: Masking rules are not applied through security_invoker views CVE-20

## "I Strongly Suspect This Website Is a Scam": Benchmarking PII Leakage and Detection without Defense in Autonomous Web Agents

DevFeed: ["I Strongly Suspect This Website Is a Scam": Benchmarking PII Leakage and Detection without Defense in Autonomous Web Agents](<https://devfeed.tech/articles/i-strongly-suspect-this-website-is-a-scam-benchmarking-pii-leakage-and-detection-without-defense-in-autonomous-web-agents-101347.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.00497>)

Author: Soham Roy, Sarthakbrata Halder, Arya Bharaty, Vaibhav Bhaskar, Yash Sinha, Dhruv Kumar, Srikant Panda, Murari Mandal

Published: 2026-10-27T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [pii](<https://devfeed.tech/topics/pii.md>), [Agentic AI](<https://devfeed.tech/topics/what-is-agentic-ai.md>), [dark patterns](<https://devfeed.tech/topics/dark-patterns.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-systems](<https://devfeed.tech/tags/agentic-systems.md>), [agents](<https://devfeed.tech/tags/agents.md>), [attack](<https://devfeed.tech/tags/attack.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [benchmarking](<https://devfeed.tech/tags/benchmarking.md>), [categories](<https://devfeed.tech/tags/categories.md>), [content](<https://devfeed.tech/tags/content.md>), [critical](<https://devfeed.tech/tags/critical.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [defense](<https://devfeed.tech/tags/defense.md>), [detection](<https://devfeed.tech/tags/detection.md>), [environments](<https://devfeed.tech/tags/environments.md>), [frontier](<https://devfeed.tech/tags/frontier.md>), [gap](<https://devfeed.tech/tags/gap.md>), [guidance](<https://devfeed.tech/tags/guidance.md>), [independent](<https://devfeed.tech/tags/independent.md>), [information](<https://devfeed.tech/tags/information.md>), [pii](<https://devfeed.tech/tags/pii.md>), [web](<https://devfeed.tech/tags/web.md>)

### AI overview

The Scammer4U benchmark finds that social-engineering sites can induce frontier web agents to submit critical-tier PII: leakage was 54-93% without privacy guidance, versus 0% on benign-twin sites. It covers 91 attacker-controlled environments and 10 benign baselines across eight attack vectors and 16 site categories. Prompt-level mitigations reduced leakage by sharply differing amounts across four model families, but did not reliably prevent submissions overall. Agents still submitted critical PII in 35.9% of sessions even when an independent LLM judge found that their reasoning flagged the site as suspicious, compared with 66.1% when no suspicion was verbalized. The authors argue this detection-action gap makes agent self-recognition an unreliable basis for defense and motivate interception of outbound submissions independent of the agent's reasoning.

### Source excerpt

arXiv:2606.00497v2 Announce Type: replace Abstract: Deceptive web content, widely instantiated across the internet and commonly known as \textit{social-engineering attacks}, manipulates autonomous web agents into submitting users' personally identifiable information (PII) to attacker-controlled endpoints. In this paper, we show that social-engineering attacks are highly effective at extracting critical-tier PII from frontier web agents, posing a severe risk to deployed agentic systems. To quantify this risk, we introduce \textbf{\textsc{Scammer4U}}, a pre-registered benchmark of 91 attacker-controlled environments and 10 benign-twin baselines, spanning 8 attack vectors and 16 site categories on an 8-axis factorial taxonomy that isolates the causal contribution of individual attack design factors. Across frontier agents, we find that critical-tier PII leakage reaches 54--93\% under no privacy guidance, compared to 0\% on benign-twin baselines, confirming that leakage is attack-attributable rather than incidental form-filling. Escalating prompt-level mitigation yields sharply model-dependent reductions across the four families and remains insufficient to reliably prevent critical PII submission at the pooled level. Most critically, we identify a detection--action gap: agents whose reasoning an independent LLM judge confirms has flagged the site as suspicious still submit critical PII in 35.9\% of sessions, versus 66.1\% when no suspicion is verbalized, a 30.2\% gap robust across all four model families. Our findings reveal that defenses conditioned on the agent's own recognition of an attack are gating on the wrong signal, motivating output-level interception of outbound submissions that operates independently of the agent's reasoning loop.

## A Causal Perspective on the Role of Training Dynamics for Interpreting Privacy Risks in Code Models

DevFeed: [A Causal Perspective on the Role of Training Dynamics for Interpreting Privacy Risks in Code Models](<https://devfeed.tech/articles/a-causal-perspective-on-the-role-of-training-dynamics-for-interpreting-privacy-risks-in-code-models-101338.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2512.07814>)

Author: Hua Yang, Alejandro Velasco, Sen Fang, Bowen Xu, Denys Poshyvanyk

Published: 2026-10-27T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>), [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Security & Privacy](<https://devfeed.tech/topics/security-privacy.md>), [pii](<https://devfeed.tech/topics/pii.md>), [LLM Techniques](<https://devfeed.tech/topics/llm-techniques.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [also](<https://devfeed.tech/tags/also.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [behaviors](<https://devfeed.tech/tags/behaviors.md>), [building](<https://devfeed.tech/tags/building.md>), [but](<https://devfeed.tech/tags/but.md>), [can](<https://devfeed.tech/tags/can.md>), [code](<https://devfeed.tech/tags/code.md>), [computing](<https://devfeed.tech/tags/computing.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [data](<https://devfeed.tech/tags/data.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [developer](<https://devfeed.tech/tags/developer.md>), [developer-productivity](<https://devfeed.tech/tags/developer-productivity.md>), [developing](<https://devfeed.tech/tags/developing.md>), [easy](<https://devfeed.tech/tags/easy.md>), [models](<https://devfeed.tech/tags/models.md>), [perspective](<https://devfeed.tech/tags/perspective.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [risks](<https://devfeed.tech/tags/risks.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

The study reports that code-model leakage varies by PII type and is associated with how readily examples are learned. The researchers built a dataset spanning PII types, fine-tuned representative models at different scales, measured training dynamics on real PII data, and used a structural causal model to estimate the effect of learnability on leakage. IP addresses, described as easier to learn, showed higher leakage; keys and passwords leaked less frequently, while ambiguous types had mixed results. The abstract presents this as causal evidence for type-dependent risk and suggests type-aware, learnability-aware defenses. These claims are limited to the study described in the abstract.

### Source excerpt

arXiv:2512.07814v3 Announce Type: replace-cross Abstract: Large language models for code (LLM4Code) have greatly improved developer productivity but also raise privacy concerns due to their reliance on open-source repositories containing abundant personally identifiable information (PII). Prior work shows that commercial models can reproduce sensitive PII, yet existing studies largely treat PII as a single category and overlook the heterogeneous risks among different types. We investigate whether distinct PII types vary in their likelihood of being learned and leaked by LLM4Code, and whether this relationship is causal. Our methodology includes building a dataset with diverse PII types, fine-tuning representative models of different scales, computing training dynamics on real PII data, and formulating a structural causal model to estimate the causal effect of learnability on leakage. Results show that leakage risks differ substantially across PII types and correlate with their training dynamics: easy-to-learn instances such as IP addresses exhibit higher leakage, while harder types such as keys and passwords leak less frequently. Ambiguous types show mixed behaviors. This work provides the first causal evidence that leakage risks are type-dependent and offers guidance for developing type-aware and learnability-aware defenses for LLM4Code.

## Veil-Then-Reveal: Protecting PII Privacy via Reversible Pseudo-Anonymization in Zero-Trust User--LLM Interaction

DevFeed: [Veil-Then-Reveal: Protecting PII Privacy via Reversible Pseudo-Anonymization in Zero-Trust User--LLM Interaction](<https://devfeed.tech/articles/veil-then-reveal-protecting-pii-privacy-via-reversible-pseudo-anonymization-in-zero-trust-user-llm-interaction-101373.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.22800>)

Author: Mohamed Elmahallawy, Mohamed Shaaban, Thomas E. Carroll

Published: 2026-10-27T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [pii](<https://devfeed.tech/topics/pii.md>), [Security & Privacy](<https://devfeed.tech/topics/security-privacy.md>), [Zero Trust](<https://devfeed.tech/topics/zero-trust.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [online privacy](<https://devfeed.tech/topics/online-privacy.md>)

Tags: [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [encryption](<https://devfeed.tech/tags/encryption.md>), [llm](<https://devfeed.tech/tags/llm.md>), [pii](<https://devfeed.tech/tags/pii.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [zero-trust](<https://devfeed.tech/tags/zero-trust.md>)

### AI overview

Veil-then-Reveal (VtR) keeps plaintext personally identifiable information on the user's device by replacing detected PII with realistic, same-type surrogates before prompts reach remote LLM providers, treated as untrusted. It uses format-preserving encryption for structured values such as phone and credit-card numbers, and a keyed-hash dictionary with a secure vault for semantic values such as names and organizations. Session-specific mappings support cross-session unlinkability, and the user restores original values in responses. A rank-class FPE strategy handles dates while preserving format and ambiguity and reducing cipher evaluations. Across three datasets with 11,322 annotated PII instances, the reported results are 97.9% detection recall, 95.2% protection of detected values, and 99.9% round trip; echoed surrogates had 100% end-to-end restoration.

### Source excerpt

arXiv:2610.22800v2 Announce Type: replace Abstract: Large language models (LLMs) increasingly process sensitive documents and personal information, yet sending prompts containing personally identifiable information (PII) to remote providers exposes data subjects and organizations to risks from logging, retention, model training, and unintended disclosure. Existing anonymization methods often mask or remove PII, while conventional encryption eliminates the lexical and structural cues needed for effective LLM processing. We present Veil-then-Reveal (VtR), a user-side PII pseudo-anonymization framework for privacy-preserving LLM interaction under a zero-trust threat model. VtR treats remote LLM providers as untrusted and sends only pseudo-anonymized prompts, keeping plaintext PII on the user. VtR first detects PII and replaces it with realistic, same-type, in-vocabulary surrogates before prompts leave the user's device, preserving their linguistic and structural properties. It combines format-preserving encryption (FPE) for structured PII, such as phone and credit-card numbers, with a keyed-hash dictionary and secure vault for semantic PII, such as names, locations, and organizations. Session-specific mappings provide cross-session unlinkability, while user-side restoration reversibly replaces surrogates with their original values in LLM responses. We further introduce a rank-class FPE strategy for dates that preserves format and ambiguity while reducing cipher evaluations. Evaluation on three datasets containing 11,322 annotated PII instances achieves 97.9% detection recall, 95.2% protection of detected values, and 99.9% round trip; 100% of echoed surrogates end-to-end round-trip restoration.

## Data governance with annotations in Amazon S3

DevFeed: [Data governance with annotations in Amazon S3](<https://devfeed.tech/articles/data-governance-with-annotations-in-amazon-s3-101933.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/architecture/data-governance-with-annotations-in-amazon-s3/>)

Author: Aditya Jain

Published: 2026-10-26T18:00:26Z

Content type: article

Language: en

Sources: [AWS Architecture Blog](<https://devfeed.tech/sources/aws-architecture-blog.md>)

Topics: [Amazon S3](<https://devfeed.tech/topics/amazon-s3.md>), [Amazon S3 Tables](<https://devfeed.tech/topics/amazon-s3-tables.md>), [Authorization](<https://devfeed.tech/topics/authorization.md>), [pii](<https://devfeed.tech/topics/pii.md>), [ai-governance](<https://devfeed.tech/topics/ai-governance.md>), [Publish-subscribe pattern](<https://devfeed.tech/topics/pubsub.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>)

Tags: [amazon-athena](<https://devfeed.tech/tags/amazon-athena.md>), [amazon-bedrock](<https://devfeed.tech/tags/amazon-bedrock.md>), [amazon-eventbridge](<https://devfeed.tech/tags/amazon-eventbridge.md>), [amazon-s3](<https://devfeed.tech/tags/amazon-s3.md>), [amazon-s3-tables](<https://devfeed.tech/tags/amazon-s3-tables.md>), [amazon-simple-storage-service-s3](<https://devfeed.tech/tags/amazon-simple-storage-service-s3.md>), [amazon-sqs](<https://devfeed.tech/tags/amazon-sqs.md>), [annotations](<https://devfeed.tech/tags/annotations.md>), [apache](<https://devfeed.tech/tags/apache.md>), [apache-iceberg](<https://devfeed.tech/tags/apache-iceberg.md>), [api](<https://devfeed.tech/tags/api.md>), [apis](<https://devfeed.tech/tags/apis.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [athena](<https://devfeed.tech/tags/athena.md>), [audit](<https://devfeed.tech/tags/audit.md>), [audit-trail](<https://devfeed.tech/tags/audit-trail.md>), [classification](<https://devfeed.tech/tags/classification.md>), [data-governance](<https://devfeed.tech/tags/data-governance.md>), [governance](<https://devfeed.tech/tags/governance.md>), [intermediate-200](<https://devfeed.tech/tags/intermediate-200.md>), [pii](<https://devfeed.tech/tags/pii.md>), [retention](<https://devfeed.tech/tags/retention.md>), [technical-how-to](<https://devfeed.tech/tags/technical-how-to.md>)

### AI overview

Amazon S3 annotations keep structured governance context on a specific object version, where it can be updated without rewriting the object and read through annotation APIs or queried at scale with an S3 Metadata Iceberg table and Athena. The pattern uses an enrichment process for new uploads, scheduled updates with an ETag condition, and S3 Batch Operations for backfill; generated labels need validation before they drive consequential actions. Key constraints shape deployment: in versioned buckets, ordinary deletes leave annotations on prior versions; Object Lock blocks annotation changes during retention; multipart copies require an opt-in to preserve annotations; annotation overwrites and deletes have no history; and SSE-C objects are unsupported. Separate read, write, and delete permissions, configure annotation replication explicitly, and account for asynchronous table refresh and annotation replication, as well as storage and request costs. Annotations suit rich, mutable context, but not relational queries, policy or lifecycle filtering on annotation values, or very high write rates to one object.

### Source excerpt

Attach governance context such as classification, retention, residency, and PII flags directly to your Amazon S3 objects with S3 annotations. This post walks through the end-to-end pattern and the handful of S3 behaviors that decide whether your governance design is correct.

## Passwordless Authentication and Identity Theft: Can Passkeys Make the Web Safer?

DevFeed: [Passwordless Authentication and Identity Theft: Can Passkeys Make the Web Safer?](<https://devfeed.tech/articles/passwordless-authentication-and-identity-theft-can-passkeys-make-the-web-safer-101472.md>)

Original publisher: [Read original article](<https://www.sitepoint.com/passwordless-authentication-and-identity-theft-can-passkeys-make-the-web-safer/>)

Author: Div John

Published: 2026-10-26T12:17:53Z

Content type: article

Language: en

Sources: [SitePoint](<https://devfeed.tech/sources/sitepoint.md>)

Topics: [Passkeys](<https://devfeed.tech/topics/passkeys.md>), [account takeover](<https://devfeed.tech/topics/account-takeover.md>), [Credential theft](<https://devfeed.tech/topics/credential-theft.md>), [Authentication](<https://devfeed.tech/topics/authentication.md>), [credential-based attacks](<https://devfeed.tech/topics/credential-based-attacks.md>), [Back end](<https://devfeed.tech/topics/backend.md>), [password reset](<https://devfeed.tech/topics/password-reset.md>), [pii](<https://devfeed.tech/topics/pii.md>), [Secure token management](<https://devfeed.tech/topics/secure-token-management.md>)

Tags: [account-takeover](<https://devfeed.tech/tags/account-takeover.md>), [authentication](<https://devfeed.tech/tags/authentication.md>), [breach](<https://devfeed.tech/tags/breach.md>), [browser](<https://devfeed.tech/tags/browser.md>), [credential-stuffing](<https://devfeed.tech/tags/credential-stuffing.md>), [identity](<https://devfeed.tech/tags/identity.md>), [identity-theft](<https://devfeed.tech/tags/identity-theft.md>), [passkeys](<https://devfeed.tech/tags/passkeys.md>), [security](<https://devfeed.tech/tags/security.md>), [web](<https://devfeed.tech/tags/web.md>)

### AI overview

Passkeys reduce account takeovers by replacing shared passwords with site-specific cryptographic key pairs: the private key stays on the user's device or synced password manager, while the site stores a public key. Domain binding blocks phishing, unique credentials defeat credential stuffing, and database theft exposes keys that cannot sign in by themselves. They do not prevent weak account recovery, malware-based session theft, leaks of personal data, or social engineering. For implementation, the excerpt advises generating challenges and user IDs on the server, setting the relying party ID to the sign-in domain, and using a maintained WebAuthn library to verify responses. Roll out gradually, allow multiple passkeys, strengthen recovery, and show users their registered credentials. Password fallback eases adoption but remains phishable; strict signature-counter checks can also lock out users with synced passkeys.

### Source excerpt

null Continue reading Passwordless Authentication and Identity Theft: Can Passkeys Make the Web Safer? on SitePoint.

## Synthetic identities leave US lenders with $2.4 billion in fraud exposure

DevFeed: [Synthetic identities leave US lenders with $2.4 billion in fraud exposure](<https://devfeed.tech/articles/synthetic-identities-leave-us-lenders-with-2-4-billion-in-fraud-exposure-99702.md>)

Original publisher: [Read original article](<https://www.helpnetsecurity.com/2026/10/26/transunion-synthetic-identity-fraud-risks-report/>)

Author: Anamarija Pogorelec

Published: 2026-10-26T05:00:59Z

Content type: news

Language: en

Sources: [Help Net Security](<https://devfeed.tech/sources/help-net-security.md>)

Topics: [fake account fraud](<https://devfeed.tech/topics/fake-account-fraud.md>), [online-fraud-prevention](<https://devfeed.tech/topics/online-fraud-prevention.md>), [account takeover](<https://devfeed.tech/topics/account-takeover.md>), [Payment Fraud](<https://devfeed.tech/topics/payment-fraud.md>), [Credential theft](<https://devfeed.tech/topics/credential-theft.md>), [pii](<https://devfeed.tech/topics/pii.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [account](<https://devfeed.tech/tags/account.md>), [ai](<https://devfeed.tech/tags/ai.md>), [attack](<https://devfeed.tech/tags/attack.md>), [authentication](<https://devfeed.tech/tags/authentication.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [digital](<https://devfeed.tech/tags/digital.md>), [financial](<https://devfeed.tech/tags/financial.md>), [financial-industry](<https://devfeed.tech/tags/financial-industry.md>), [fraud](<https://devfeed.tech/tags/fraud.md>), [identity](<https://devfeed.tech/tags/identity.md>), [identity-protection](<https://devfeed.tech/tags/identity-protection.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [manipulate](<https://devfeed.tech/tags/manipulate.md>), [money](<https://devfeed.tech/tags/money.md>), [network](<https://devfeed.tech/tags/network.md>), [news](<https://devfeed.tech/tags/news.md>), [personal](<https://devfeed.tech/tags/personal.md>), [report](<https://devfeed.tech/tags/report.md>), [risks](<https://devfeed.tech/tags/risks.md>), [scams](<https://devfeed.tech/tags/scams.md>), [security](<https://devfeed.tech/tags/security.md>), [synthetic](<https://devfeed.tech/tags/synthetic.md>), [transactions](<https://devfeed.tech/tags/transactions.md>), [transunion](<https://devfeed.tech/tags/transunion.md>), [trust](<https://devfeed.tech/tags/trust.md>), [us](<https://devfeed.tech/tags/us.md>)

### AI overview

TransUnion reports that 81% of suspected digital fraud in its network in the first half of 2026 occurred during logins or financial transactions, showing that identity checks at onboarding do not protect the full customer relationship. The excerpt describes impersonation scams that persuade customers to share codes, reset credentials or redirect payments, and synthetic identities built from exposed personal data and AI that can pass verification and be reused across services. US lenders had an estimated $2.4 billion in synthetic identity fraud risk exposure at the end of June 2026, including $1.5 billion in auto lending; these figures are potential losses, not losses already incurred. The report also notes attackers target account recovery and secondary authentication, while fraud patterns vary by region.

### Source excerpt

Criminals are exploiting established customer accounts and familiar brands to steal money and personal information, according to TransUnion's Fraud in the Age of Continuous Trust report. Eighty-one percent of suspected digital fraud in its network during the first half of 2026 occurred during account logins and financial transactions, highlighting risks that persist throughout a customer relationship. Verifying someone's identity does not establish whether every subsequent interaction can be trusted. Scammers can manipulate legitimate customers into ... More -> The post Synthetic identities leave US lenders with $2.4 billion in fraud exposure appeared first on Help Net Security.

## CVE-2026-55826 - Customer Order Data Exposed via Unauthenticated gtm\_server\_side\_order\_id Cookie Manipulation in GTM Server Side WordPress Plugin

DevFeed: [CVE-2026-55826 - Customer Order Data Exposed via Unauthenticated gtm\_server\_side\_order\_id Cookie Manipulation in GTM Server Side WordPress Plugin](<https://devfeed.tech/articles/cve-2026-55826-customer-order-data-exposed-via-unauthenticated-gtm-server-side-order-id-cookie-manipulation-in-gtm-server-side-wordpress-plugin-98328.md>)

Original publisher: [Read original article](<https://cvefeed.io/vuln/detail/CVE-2026-55826>)

Author: CVEFEED.io; github.com

Published: 2026-10-23T14:17:54Z

Content type: news

Language: en

Sources: [Latest High/Critical Vulnerabilitiy Feed](<https://devfeed.tech/sources/latest-high-critical-vulnerabilitiy-feed.md>)

Topics: [pii](<https://devfeed.tech/topics/pii.md>), [WordPress](<https://devfeed.tech/topics/wordpress.md>)

Tags: [2-0-0](<https://devfeed.tech/tags/2-0-0.md>), [2026](<https://devfeed.tech/tags/2026.md>), [allows](<https://devfeed.tech/tags/allows.md>), [customer](<https://devfeed.tech/tags/customer.md>), [cve](<https://devfeed.tech/tags/cve.md>), [cve-2026-55826](<https://devfeed.tech/tags/cve-2026-55826.md>), [data](<https://devfeed.tech/tags/data.md>), [event](<https://devfeed.tech/tags/event.md>), [exposed](<https://devfeed.tech/tags/exposed.md>), [full](<https://devfeed.tech/tags/full.md>), [gtm](<https://devfeed.tech/tags/gtm.md>), [information](<https://devfeed.tech/tags/information.md>), [issue](<https://devfeed.tech/tags/issue.md>), [javascript](<https://devfeed.tech/tags/javascript.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [order](<https://devfeed.tech/tags/order.md>), [plugin](<https://devfeed.tech/tags/plugin.md>), [server](<https://devfeed.tech/tags/server.md>), [wordpress](<https://devfeed.tech/tags/wordpress.md>), [wordpress-plugin](<https://devfeed.tech/tags/wordpress-plugin.md>)

### AI overview

In GTM Server Side for WordPress, versions 2.0.0 through before 2.1.48 have a WooCommerce purchase tracking flaw: an unauthenticated attacker can manipulate the `gtm_server_side_order_id` cookie to cause another customer's order data, including personally identifiable information, to be rendered in the page's JavaScript dataLayer. The issue is patched in version 2.1.48. The source reports severity 8.7 (HIGH), but provides no further affected-product records or technical remediation details.

### Source excerpt

CVE ID :CVE-2026-55826 Published : Oct. 23, 2026, 2:17 p.m. | 5 hours, 25 minutes ago Description :The GTM Server Side plugin by stape.io implements server-side tracking on WordPress websites. From version 2.0.0 to before version 2.1.48, the WooCommerce purchase event tracking mechanism allows an unauthenticated attacker to trigger the rendering of another customer's full order data -- including personally identifiable information (PII) -- into the page's JavaScript dataLayer. This issue has been patched in version 2.1.48. Severity: 8.7 | HIGH Visit the link for more details, such as CVSS details, affected products, timeline, and more...

[Next page](<https://devfeed.tech/topics/pii.md?cursor=WyIyMDI2LTEwLTIzVDE0OjE3OjU0KzAwOjAwIiwgImQ3NWE0NzU0LWY5NjAtNDNlYi04YzY2LWZiNzZlNWU1OTQ0NSJd>)