<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Prashant on Medium]]></title>
        <description><![CDATA[Stories by Prashant on Medium]]></description>
        <link>https://medium.com/@learningisfun?source=rss-8e485a9bf87a------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*keoK60oj-ZbLCk7eG0BA7w.png</url>
            <title>Stories by Prashant on Medium</title>
            <link>https://medium.com/@learningisfun?source=rss-8e485a9bf87a------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 12:07:24 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@learningisfun/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Zero Trust for AI Agents]]></title>
            <link>https://learningisfun.medium.com/zero-trust-for-ai-agents-db6f1336986b?source=rss-8e485a9bf87a------2</link>
            <guid isPermaLink="false">https://medium.com/p/db6f1336986b</guid>
            <dc:creator><![CDATA[Prashant]]></dc:creator>
            <pubDate>Thu, 04 Jun 2026 00:37:01 GMT</pubDate>
            <atom:updated>2026-06-04T00:37:01.783Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*l8iRsggssbWRH4NXYWejKQ.png" /></figure><p>Anthropic recently published a white paper entitled “Zero Trust for AI Agents,” which is available for download at <a href="https://cdn.prod.website-files.com/6889473510b50328dbb70ae6/6a1611a04085d7cd3dadc924_Claude-eBook-Zero-Trust-for-AI-Agents-05182026.pdf">https://cdn.prod.website-files.com/6889473510b50328dbb70ae6/6a1611a04085d7cd3dadc924_Claude-eBook-Zero-Trust-for-AI-Agents-05182026.pdf</a></p><p>The white paper describes the core principle that every cybersecurity professional is aware of Zero Trust — Never trust, always verify, assume breach. This must be extended to autonomous AI agents, not just human identities and traditional software. Agentic AI introduces fundamentally new attack surfaces: autonomous multi-step execution, tool access, context persistence, and multi-agent coordination.A tiered security model (Foundation → Enterprise → Advanced) provides a roadmap for incrementally hardening agent deployments. The key design test for any control: does it make the attack <em>impossible</em>, or merely <em>tedious</em>? AI-accelerated attackers can grind through tedious friction at near-zero cost. Defensive operations must themselves run at machine speed with humans owning decisions, not bookkeeping.</p><p>It outlines a comprehensive security framework for deploying autonomous AI agents in enterprise environments, rooted in Zero Trust principles. It highlights that traditional, perimeter-based cybersecurity defenses are insufficient against modern, AI-accelerated threats, which compress the timeline between vulnerability and exploit from months to hours at a marginal cost. Because AI agents execute multi-step operations and select tools autonomously, they introduce unique vulnerabilities that legacy access controls cannot mitigate. OWASP identifies five primary threat categories for agentic systems: prompt injection, tool/resource hijacking, identity/privilege abuse, memory/context poisoning, and supply chain risks.</p><p>The narrative introduces a three-tiered security capability model, Foundation, Enterprise, and Advanced designed to establish secure baselines, mitigate trending threats, and run defensive operations at machine speed. The framework emphasizes that the minimum viable security “floor” has been raised; basic friction-based controls are no longer acceptable, demanding a shift toward cryptographic identities, short-lived tokens, and strict sandboxing even at the Foundation level.</p><p>Ultimately, the text concludes that the most resilient organizations will not necessarily be those with the most advanced AI models. Instead, they will be the organizations whose fundamentals are strong enough to prevent AI-assisted scanning from finding bugs, and whose agentic systems are architected with an “assume breach” mindset from day one.</p><h3>These are the core Ideas that are defined by Anthropic</h3><ul><li><strong>Zero Trust and “Least Agency”:</strong> Extending the traditional principle of least privilege to autonomous systems, “least agency” restricts exactly what an AI agent’s tools can do, where they can do it, and how often. This operates under the assumption that a breach is inevitable and aims to minimize the potential blast radius. “Least agency” (OWASP’s extension of least privilege) restricts not just <em>what</em> agents can access, but <em>how often</em> and <em>in what context</em> e.g., a database tool gets read-only access, an email tool gets no send/delete rights. The US requires all federal agencies to adopt Zero Trust by 2027; HIPAA, FINRA, GDPR, FedRAMP, and the EU AI Act already impose aligned requirements.</li><li><strong>The “Impossible vs. Tedious” Design Test:</strong> Security controls must create hard barriers rather than merely inconvenient friction. Controls like rate limits, extra pivot hops, or SMS-based MFA fail against AI attackers, who possess unlimited patience and near-zero per-attempt costs. Effective controls rely on hardware-bound credentials and cryptographic identity.</li><li><strong>Agent-Specific Threat Landscape:</strong> Agents face unique vulnerabilities beyond traditional software flaws. Critical threats include indirect prompt injections hidden in processed external data, tool poisoning via falsified capabilities, unscoped privilege inheritance, and memory poisoning that alters long-term agent behavior.</li><li><strong>Three-Tiered Implementation Framework:</strong> A tiered security model (Foundation → Enterprise → Advanced) provides a roadmap for incrementally hardening agent deployments. Security maturity is structured into three levels. The <strong>Foundation</strong> tier represents minimum viable security (e.g., short-lived tokens, RBAC). The <strong>Enterprise</strong> tier reflects standard practices (e.g., sandboxed execution, ABAC). The <strong>Advanced</strong> tier covers stringent requirements (e.g., hardware isolation, continuous authorization).</li><li><strong>Autonomous Defensive Operations:</strong> Defenders must match the machine speed of autonomous threats. Organizations must automate evidence collection, enrichment, and triage such as placing a triage AI model at the front of the alert queue keeping human analysts focused exclusively on making critical containment and disclosure decisions.</li></ul><h3>Key Insights &amp; Takeaways</h3><ul><li><strong>Indirect Injections are Insidious:</strong> Large Language Models (LLMs) cannot reliably distinguish between informational context and actionable instructions. This makes indirect prompt injections embedded in seemingly benign web pages or emails a massive threat, as the payload is executed without the user ever seeing it. Injecting just 250 malicious documents can backdoor LLMs with 600M–13B parameters, and these backdoors survive safety training including RLHF. Microsoft’s Spotlighting technique reduces indirect prompt injection success from &gt;50% to &lt;2% by clearly delimiting untrusted content for the model.</li><li><strong>Static Credentials are Compromised by Default:</strong> API keys and shared service passwords are no longer viable entry points, even at the Foundation level of security. Attackers utilizing model-assisted code analysis will find them immediately; security must rely on short-lived, identity-provider-issued tokens.</li><li><strong>AI-BOMs are Essential for Supply Chains:</strong> Because AI ecosystems load external tools and personas dynamically, supply chain risks are amplified. Organizations must treat model components with the same rigor as code dependencies by implementing an AI Bill of Materials (AI-BOM) to track provenance and fine-tuning data.</li><li><strong>Dwell Time and Coverage are Priority Metrics:</strong> Before investing in other detection methods, organizations should instrument dwell time (anomaly occurrence to human awareness) and coverage (fraction of alerts investigated). AI-assisted automation provides the greatest leverage to improve these specific metrics.</li><li><strong>Tool Chaining Bypasses Traditional Monitoring:</strong> Attackers can trick agents into combining legitimate tools in harmful sequences (e.g., linking an internal CRM tool to an external email tool to exfiltrate data). Because these commands execute through trusted binaries, host-centric monitoring detects no malware.</li></ul><h3>Action Items &amp; Next Steps</h3><ul><li><strong>Zero Trust Controls</strong> — Apply Zero Trust controls to defensive agents (agentic SOAR) with the same rigor as offensive-facing agents.</li><li><strong>Deprecate Static Credentials:</strong> Replace all static API keys, embedded credentials, and shared service-account passwords with short-lived, identity-provider-issued tokens or hardware-bound credentials for production workloads.</li><li><strong>Assign Cryptographic Identities:</strong> Provision unique, cryptographically rooted identifiers for every agent instance to ensure traceability, enable accurate auditing, and enforce identity-based isolation.</li><li><strong>Establish AI Governance:</strong> Form a cross-functional AI governance committee consisting of security, legal, compliance, and business stakeholders. Document acceptable use, establish incident response procedures, and define approval processes for new agent deployments. Define explicit approved/prohibited actions, escalation triggers, and scope limits per agent before deployment. Implement input isolation (spotlighting) and output filtering for PII/credentials on all agents processing external data</li><li><strong>Implement Automated Alert Triage:</strong> Deploy a triage AI model with read-only SIEM access to the front of the alert queue. Start by wiring it into one noisy rule with a known-high false positive rate to produce structured dispositions, and measure its agreement with human reviewers before expanding.</li><li><strong>Audit Dependency Trees:</strong> Point a frontier model at existing lockfiles to audit for redundant dependencies (e.g., multiple HTTP clients) and execute consolidation efforts to reduce the attack surface. Use OpenSSF Scorecard to evaluate dependency health automatically; audit the dependency tree for redundant libraries.</li><li><strong>Map Defenses to MITRE ATT&amp;CK:</strong> Run tests using the open-source Atomic Red Team library to verify what existing logging actually detects, prioritizing coverage for lateral movement and credential access. Instrument two metrics before anything else: dwell time (anomaly to human awareness) and alert coverage (fraction actually investigated).Map detection coverage against MITRE ATT&amp;CK; prioritize lateral movement and credential access techniques.</li><li><strong>Pre-Authorize Emergency Procedures:</strong> Define and practice the authorization path for emergency containment actions, such as taking a service offline or rotating a credential, so that decision-making is not improvised during an active incident.</li><li><strong>Conduct Multi-Incident Tabletop Exercises:</strong> Rehearse a tabletop exercise where five critical vulnerabilities hit in the same week, rather than just one, to ensure intake and remediation tracking can scale appropriately. Run tabletop exercises simulating five simultaneous incidents, not one and pre-authorize emergency containment actions before an incident occurs.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=db6f1336986b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Verify SRV records after DC pro]]></title>
            <link>https://learningisfun.medium.com/verify-srv-records-after-dc-pro-22f9aa2cb0af?source=rss-8e485a9bf87a------2</link>
            <guid isPermaLink="false">https://medium.com/p/22f9aa2cb0af</guid>
            <category><![CDATA[active-directory]]></category>
            <dc:creator><![CDATA[Prashant]]></dc:creator>
            <pubDate>Wed, 20 Aug 2025 02:09:06 GMT</pubDate>
            <atom:updated>2025-08-20T02:09:06.532Z</atom:updated>
            <content:encoded><![CDATA[<h3>Verify SRV records after DC promotion</h3><p>To configure a Windows Server as a Domain Controller, the DCPROMO command is executed. Once the Domain Controller is set up, it&#39;s essential to verify that all necessary SRV records have been created successfully. We can utilize the nslookup command to view these records for the newly created Domain Controller.</p><p><strong>Using NSLookup to Verify SRV Records:</strong></p><ol><li>On your DNS server, open a <strong>Command Prompt</strong> and type nslookup.</li><li>Enter set type=all and press <strong>ENTER</strong>.</li><li>Type _ldap._tcp.dc._msdcs.&lt;domainname&gt; (replace &lt;domainname&gt; with your actual domain name) and press <strong>ENTER</strong>.</li></ol><p><strong>Other way to verify SRV records is utilizing the DCDiag command :</strong></p><p>Use the dcdiag command to verify all SRV records for a Domain Controller:</p><p>dcdiag /s:&lt;domain_name&gt; /test:dns /dnsrecordregistration /v</p><ul><li>Replace &lt;domain_name&gt; with your actual domain name.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=22f9aa2cb0af" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Distributed File System (DFS)]]></title>
            <link>https://learningisfun.medium.com/distributed-file-system-dfs-d5c0afb72c48?source=rss-8e485a9bf87a------2</link>
            <guid isPermaLink="false">https://medium.com/p/d5c0afb72c48</guid>
            <category><![CDATA[active-directory]]></category>
            <category><![CDATA[dfs]]></category>
            <dc:creator><![CDATA[Prashant]]></dc:creator>
            <pubDate>Wed, 20 Aug 2025 02:05:49 GMT</pubDate>
            <atom:updated>2025-08-20T02:05:49.834Z</atom:updated>
            <content:encoded><![CDATA[<p>DFS (Distributed File System) enables the creation of a network file share that encompasses multiple file shares. By consolidating all shares under a single name, known as a DFS namespace, users can access all shares without having to browse individual servers. This simplifies the user experience, as they only need to remember a single namespace to access all available shares on the network or domain. Additionally, DFS facilitates fault tolerance, load sharing, and replication between systems, making it an ideal solution for companies with multiple file servers.</p><p><strong>DFS Technologies: Namespace and Replication</strong></p><p>DFS consists of two distinct technologies:</p><ol><li><strong>DFS Namespace</strong>: Allows administrators to group shared folders from different servers into a virtual tree of folders, presented as a single namespace.</li><li><strong>DFS Replication (DFSR)</strong>: Utilizes a multi-master replication engine to distribute identical data across multiple servers.</li></ol><p><strong>Key Considerations Before Implementing DFS</strong></p><p>Before deploying DFS, consider the following constraints:</p><ul><li><strong>Suitability</strong>: Ideal for consolidating shares under a single namespace, but not recommended for environments with frequent data changes or simultaneous updates to the same data.</li><li><strong>File Check-out</strong>: For scenarios with multiple users modifying the same file, use the file check-out feature of Windows SharePoint Services to ensure exclusive access.</li><li><strong>Replication</strong>: DFSR replicates data after files are closed, without real-time syncing or file locking across member servers.</li><li><strong>Incompatibility</strong>: Not suitable for Exchange Servers, SQL servers, or similar applications; recommended for file-based resources that are opened and then closed.</li></ul><p><strong>DFS Namespace: Benefits and Example</strong></p><p>The DFS namespace technology offers:</p><ul><li><strong>Increased Data Availability</strong></li><li><strong>Load Sharing</strong></li><li><strong>Simplified Data Migration</strong></li></ul><p><strong>Practical Example 1: Consolidating Shares</strong></p><p>Suppose you have three shares:</p><ul><li>\\server1\Finance</li><li>\\server2\Marketing</li><li>\\server3\HR</li></ul><p>Create a single namespace, \\citruspk.com\dfsroot, to provide users with a unified access point. Users can then browse to \\citruspk.com\dfsroot and access individual folders (Finance, Marketing, HR) based on their assigned permissions.</p><p><strong>Practical Example 2: Main and Branch Office Scenario</strong></p><ul><li><strong>Scenario</strong>: Same data is spread across two locations (India and US servers).</li><li><strong>Solution</strong>: Create a DFS replica set to keep files synchronized between servers.</li><li><strong>Benefits</strong>:</li><li>Automatic redirection to the closest server based on the user’s location (using AD site feature).</li><li>Seamless access to data, even if one server is unavailable.</li><li>Users from each location always connect to their respective local server.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/369/0*ZvlFAuKXavt0Ql2K" /></figure><p><strong>DFS Access Flow and Replication Mechanism</strong></p><p><strong>Access Flow:</strong></p><ol><li><strong>User Request</strong>: A user attempts to access a folder within the namespace, e.g., \\citruspk.com\dfsroot.</li><li><strong>Client-Server Interaction</strong>: The client computer contacts a DFS server.</li><li><strong>Referral Response</strong>: The DFS server sends a referral to the client computer, containing a list of servers that host the shared folders (known as “folder targets” in DFS terminology).</li><li><strong>Site-Aware Redirection</strong>: Based on the user’s location, the client is redirected to the nearest server hosting the shared data. For example:</li></ol><ul><li>A US-based user is connected to the US server, as it resides within their local site.</li><li>This site-aware functionality is made possible by DFS’s integration with Active Directory.</li></ul><ol><li><strong>Fallback to Remote Server</strong>: If the local target is unavailable, the client is redirected to a remote server hosting the same data.</li><li><strong>Consistent User Experience</strong>: To ensure seamless access, all network shares must have identical data, regardless of the target server connected to.</li></ol><p><strong>DFS Replication (DFSR) Mechanism:</strong></p><ul><li><strong>Multi-master Replication Engine</strong>: DFSR employs a multi-master replication engine, similar to Active Directory replication.</li><li><strong>Replication Scheduling</strong>: Supports scheduling replication to control when data is synchronized between servers.</li><li><strong>Bandwidth Throttling</strong>: Allows for bandwidth throttling to manage network utilization.</li><li><strong>Remote Differential Compression (RDC)</strong>: A compression protocol used for replication, which:</li><li>Compresses data during replication, reducing bandwidth requirements.</li><li>Only replicates the “delta” (changes) in files, rather than the entire file.</li><li>Detects insertions, removals, and rearrangements of data within files.</li><li>Features “cross-file RDC” to minimize bandwidth usage when replicating new files.</li><li><strong>Native Replication Protocol</strong>: DFS utilizes its native RDC protocol for replication, eliminating the need for additional replication software (e.g., SAN/NAS-based tools).</li></ul><p><strong>Prerequisites for Setting up DFS and DFS-R</strong></p><p>To successfully configure Distributed File System (DFS) and DFS Replication (DFS-R), ensure the following requirements are met:</p><p><strong>Active Directory Schema Version</strong>:</p><ul><li>The Active Directory schema must be extended to at least one of the following versions:</li><li>Windows 2012 and above</li></ul><p><strong>Forest Membership</strong>:</p><ul><li>All file servers participating in a replication group must belong to the <strong>same</strong> Active Directory forest.</li><li><strong>Note:</strong> Replication across different forests is not supported.</li></ul><p><strong>DFS Replication Installation</strong>:</p><ul><li>DFS Replication must be installed on <strong>all</strong> member servers in the replication group.</li></ul><p><strong>Server Configuration</strong>:</p><p><strong>Server Core</strong> installations are <strong>not supported</strong> for managing DFS, as the DFS Management snap-in requires a full server installation.</p><p><strong>Antivirus Software Compatibility</strong>:</p><p>Ensure that your antivirus software is <strong>compatible with DFS Replication</strong>, allowing DFS traffic to pass through without interference.</p><p><strong>Active Directory Site Configuration</strong>:</p><p>Properly <strong>configure AD sites and subnets</strong>, as DFS relies on this information to identify the nearest DFS replica for clients.</p><h3>1. Domain-Based Namespace</h3><p>This configuration offers enhanced <strong>scalability</strong> and <strong>fault tolerance</strong>, making it the recommended choice. All DFS-related settings are stored in the Active Directory database. If an Active Directory infrastructure is already in place, opting for a domain-based DFS namespace is advisable.</p><h3>2. Standalone Namespace</h3><p>In this configuration, the DFS setup is stored in the registry of the DFS server itself.</p><p><strong>How Clients/Users Connect to the Nearest Shared Folder/Replica</strong></p><p>When a user attempts to access a Distributed File System (DFS) folder, the following process unfolds behind the scenes:</p><ol><li><strong>Initial Query</strong>: DFS-aware clients utilize the Netlogon service to query Active Directory (AD) for the names of namespace servers associated with the DFS folder.</li><li><strong>Namespace Server Selection</strong>:</li></ol><ul><li>Based on AD site information, the Netlogon service attempts to establish contact with the closest available namespace server within the client’s site.</li><li>If no namespace server is available in the client’s site, the client automatically selects the next closest offsite server.</li></ul><p><strong>Note:</strong> This feature requires the <strong>DFS Site Costing</strong> feature to be enabled by running the command DFSutil /SiteCosting (which must be turned ON).</p><p><strong>Folder Access and Referral</strong>:</p><ul><li>Once connected to the namespace server, the client receives the folder structure, allowing the user to select a specific folder.</li><li>The namespace server provides a referral to the network share, which the client uses to connect to the destination server hosting the actual data. If multiple targets are available for the selected folder, the client again leverages AD site information to choose the closest share.</li></ul><p><strong>Storage of DFS Information in Active Directory</strong>: DFS configuration data is stored in AD as Binary Large OBjects (BLOBs) within the domain partition at the following location: CN=Dfs-Configuration,CN=System,DC=&lt;domain&gt;,DC=&lt;domain name&gt;.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d5c0afb72c48" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Promoting DC to Global Catalog Server]]></title>
            <link>https://learningisfun.medium.com/promoting-dc-to-global-catalog-server-13d1c0028968?source=rss-8e485a9bf87a------2</link>
            <guid isPermaLink="false">https://medium.com/p/13d1c0028968</guid>
            <category><![CDATA[global-catalog]]></category>
            <category><![CDATA[active-directory]]></category>
            <dc:creator><![CDATA[Prashant]]></dc:creator>
            <pubDate>Wed, 20 Aug 2025 01:53:45 GMT</pubDate>
            <atom:updated>2025-08-20T01:53:45.691Z</atom:updated>
            <content:encoded><![CDATA[<p>When working with Active Directory, you may be instructed to promote a Domain Controller (DC) to a Global Catalog (GC) server. However, if your Active Directory database is large (e.g., approximately 13 GB), replication of all partitions to the new GC server can take several hours (around 8–9 hours). The crucial question is: how do you verify that replication is in progress and proceeding as planned?</p><p><strong>Monitoring Replication Progress</strong></p><ol><li><strong>NTDS.DIT File Size</strong>: Periodically check (every 15 minutes) the size of the NTDS.DIT file, which should increase steadily until replication is complete.</li><li><strong>Directory Service Event Log</strong>: Inspect the Event Viewer for specific events:</li></ol><ul><li>Event IDs 1927, 1110, 1128, and 1578 indicate replication progress.</li><li>Event ID 1119 confirms successful promotion to a GC server.</li></ul><p><strong>Understanding Event ID 1578</strong></p><p>Event ID 1578 may appear every 30 minutes, potentially causing concern. This informational event indicates that promotion is delayed due to unmet directory partition occupancy requirements.</p><p><strong>Event ID 1578 Contents</strong></p><ul><li><strong>Occupancy Requirement Level</strong>: 6</li><li><strong>Domain Controller Level</strong>: 1</li><li><strong>Registry Key Value</strong>: HKeyLocalMachine\System\CurrentControlSet\Services\NTDS\Parameters\Global Catalog Partition Occupancy</li></ul><p><strong>Registry Key Value Explanation</strong></p><ul><li>Defines the directory partition occupancy requirement level:</li><li>Levels 0–6 specify increasing requirements for read-only and fully synchronized directory partitions within the site and forest.</li><li><em>Higher occupancy requirement levels include the lower levels. The levels are defined as follows:</em></li></ul><p><em>(0) Indicates no occupancy requirement.</em><br><em>(1) Indicates at least one read-only directory partition in the site has been added by the Knowledge Consistency Checker (KCC).</em><br><em>(2) Indicates at least one directory partition in the site has been fully synchronized.</em><br><em>(3) Indicates all read-only directory partitions in the site have been added by the KCC (at least one has been synchronized).</em><br><em>(4) Indicates all directory partitions in the site have been fully synchronized.</em><br><em>(5) Indicates all read-only directory partitions in the forest have been added by the KCC (at least one has been synchronized).</em><br><em>(6) Indicates all directory partitions in the forest have been fully synchronized.</em></p><p><strong>The Missing Registry Value Conundrum</strong></p><p>If you navigate to the specified registry path (HKLM\System\CurrentControlSet\Services\NTDS\Parameters) and find that the &quot;Global Catalog Partition Occupancy&quot; value is absent, do not panic. According to Microsoft documentation, this value was required for Exchange 2000 environments. Before adding this value, ensure:</p><ol><li><strong>No Replication Errors</strong>: Run repadmin /replsum and resolve any errors.</li><li><strong>Domain Controller Health</strong>: Run dcdiag /s: DC Name /v /f:path of log file to check for issues.</li><li><strong>NTDS.DIT Size Increase</strong>: Verify the constant growth of the NTDS.DIT file size (e.g., in C:\Windows\NTDS).</li></ol><p><strong>If NTDS.DIT Size is Increasing…</strong></p><p>There is no need to configure the “Global Catalog Partition Occupancy” registry value. The personal experience confirms that successfully promoting a Windows 2003 R2 DC to a GC server is possible without adding this value. The appearance of Event ID 1119 in the DS event log will confirm the DC’s new role as a GC server.</p><p><strong>Troubleshooting and Diagnostic Commands</strong></p><ul><li>nltest /dclist:Domain Name: List all domain controllers in a domain, including their site names.</li><li>nltest /dsgetdc:Domain Name /server:Domain Controller Name: List all roles assigned to a domain controller, including the GC role (under Flags).</li><li>Repadmin /kcc: Force the KCC to recalculate replication topology on a DC immediately.</li><li>Repadmin /replsum: Check for replication errors with replication partners (replication summary).</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=13d1c0028968" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[PDC Emulator FSMO Role in an Active Directory Forest]]></title>
            <link>https://learningisfun.medium.com/pdc-emulator-fsmo-role-in-an-active-directory-forest-6e99bb9c0820?source=rss-8e485a9bf87a------2</link>
            <guid isPermaLink="false">https://medium.com/p/6e99bb9c0820</guid>
            <category><![CDATA[active-directory]]></category>
            <category><![CDATA[fsmo]]></category>
            <dc:creator><![CDATA[Prashant]]></dc:creator>
            <pubDate>Wed, 20 Aug 2025 01:44:38 GMT</pubDate>
            <atom:updated>2025-08-20T01:44:38.971Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Understanding the PDC Emulator FSMO Role in an Active Directory Forest</strong></p><p>We are all familiar with the five FSMO (Flexible Single Master Operations) roles available in an Active Directory (AD) forest, comprising three domain-wide roles and two forest-wide roles. In this discussion, we will delve into the most critical role, the PDC Emulator, in detail.</p><p><strong>Time Synchronization Fundamentals</strong></p><ul><li>Client machines synchronize their time with the Domain Controller (DC) that authenticates them.</li><li>All Domain Controllers in a domain synchronize their time with the PDC Emulator in domain.</li><li>The PDC Emulator in child domains synchronizes its time with a Domain Controller in the parent domain.</li><li>Ideally, the PDC Emulator should be configured to synchronize with an external time server to ensure accurate timekeeping. If not, the time set on the PDC Emulator is applied by default.</li><li>Notably, a client machine can still authenticate with a Domain Controller even with a 5-minute time difference (skew). However, Kerberos protocol relies heavily on accurate time; incorrect time will prevent authentication. To immediately synchronize a client machine’s time, run the following command:</li></ul><pre>w32tm /resync /nowait</pre><ul><li>Domain clients running Windows XP/.NET automatically synchronize their clocks with the PDC Emulator’s time upon startup and authentication. The PDC Emulator, in turn, synchronizes its clock with the parent or forest root domain’s PDC Emulator. It is recommended that the forest root domain’s PDC Emulator synchronizes with an external time server(s), although it can be left as is.</li></ul><p><strong>Key Functions of the PDC Emulator</strong></p><ol><li><strong>Time Synchronization</strong>: Acts as the time synchronizer for all Domain Controllers.</li><li><strong>Password Management</strong>: Password changes and account lockouts are always replicated to the PDC Emulator from Domain Controllers.</li><li><strong>Group Policy Management</strong>: The PDC Emulator must be available when creating or modifying Group Policies.</li></ol><p><strong>Impact of PDC Emulator Downtime</strong></p><p>From a client machine perspective, the first two functions are crucial. Consequently, if the PDC Emulator is unavailable for an extended period, the impact will be immediately noticeable, more so than with other FSMO roles.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6e99bb9c0820" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Zero Budget, Unlocked Intelligence: Free LLMs for Innovation]]></title>
            <link>https://learningisfun.medium.com/zero-budget-unlocked-intelligence-free-llms-for-innovation-dfe5f74435a4?source=rss-8e485a9bf87a------2</link>
            <guid isPermaLink="false">https://medium.com/p/dfe5f74435a4</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Prashant]]></dc:creator>
            <pubDate>Fri, 08 Aug 2025 21:32:17 GMT</pubDate>
            <atom:updated>2025-08-09T12:19:52.410Z</atom:updated>
            <content:encoded><![CDATA[<p>November 30, 2022, marked a pivotal moment in history — the day OpenAI’s ChatGPT and Large Language Model (LLM) burst into the mainstream consciousness. Though Artificial Intelligence (AI), Machine Learning, and Data Science have been quietly evolving for decades, it wasn’t until AI began transforming the job market, much like the Industrial Revolution (1760–1840) did with manufacturing, that the world took notice. Today, tech giants are in a heated competition, each touting their own LLMs. But behind the scenes, one company powers them all: Nvidia, the GPU manufacturer turned LLM chip powerhouse.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/982/0*QOW2Ff0c3OGxFAuG.png" /></figure><p><strong>The Data Crunching Backbone: Why Nvidia Stands Tall</strong></p><p>These Large Language Models aren’t just intelligent; they’re also incredibly hungry — for data. To process vast datasets, they require immense computational might. This is where Nvidia’s specialized chips come into play, empowering LLMs to learn, adapt, and generate human-like responses at unprecedented scales.</p><p><strong>The Free LLM Paradox: Can You Have Your Cake and Eat It Too?</strong></p><p>While each LLM offers a free version, limitations abound. But what if you could tap into the collective might of the latest LLMs from Deepseek, Llama, Mistral, Phi and more without spending a dime? Enter Nvidia’s hidden gem: <strong>Models</strong> (<a href="https://build.nvidia.com/models">https://build.nvidia.com/models</a>). This platform lets you experiment with top-tier models that are open-source, reaping all the benefits of these generative AI powerhouses — minus one thing: the sleek user interface that typically comes with them.</p><p><strong>Unleashing LLM Potential for Free: What to Expect on Nvidia’s Models Platform</strong></p><ul><li><strong>Access to Elite Models</strong>: Experience the capabilities of various models such as Llama from Meta, , deepseek models from deepseek, phi3 models from Microsoft and others without the cost.</li><li><strong>Limitless Experimentation</strong>: Push the boundaries of what these LLMs can do, unrestricted by typical free-tier limitations.</li><li><strong>The Trade-Off: A Utilitarian Interface</strong>: While functionality is paramount, be prepared for a more straightforward, less polished user experience.</li></ul><p><strong>Embracing the Future, for Free</strong></p><p>In the midst of the LLM frenzy, Nvidia’s Models platform stands as a beacon for innovators, learners, and the curious. Dive into the world of Large Language Models without financial barriers, and discover the vast potential of AI — courtesy of Nvidia’s pioneering spirit.</p><p><strong>How Can I Unlock the Power of Nvidia’s Models for Free?</strong></p><p>Head over to the Nvidia Models website:</p><p><a href="https://build.nvidia.com/models">Try NVIDIA NIM APIs</a></p><p>Once you’re on the site, you’ll find a treasure trove of top-tier models at your fingertips. Browse, explore, and get ready to unleash the full potential of these Large Language Models — all without spending a dime!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PepirkreoAYdsItE.png" /></figure><p>Depending on the model or resource, you might need to create a free account or sign in with existing Nvidia credentials. Don’t worry, it’s a quick hurdle!</p><p>Since you’re accessing these models directly, the UI might differ from what you’re used to. Take a minute to get comfy with the layout, and you’ll be generating like a pro in no time!</p><p>The top bar allows you to sort the models by Publisher, Use cases or the NIM Type. The NIM type means means how mature the model is. The NIM stands for <a href="https://www.google.com/search?sca_esv=733e742818f46c91&amp;cs=0&amp;q=NVIDIA+Inference+Microservices+%28NIM%29&amp;sa=X&amp;ved=2ahUKEwjeifjQwPOOAxV_hYkEHf7aN78QxccNegQIAhAB&amp;mstk=AUtExfDbdtcalQn471Go2eFAzfy46r2UVfkvBH5G40SfLKS5mpPf1NtuIIQFoADL78-s3BBIDKRai9O3GQN5txMAD2cB0O3tf1WLO5Xx-G_WedLA6O8BG4K-VIVUdiRclaRvNYA&amp;csui=3">NVIDIA Inference Microservices (NIM)</a>. There are two types that are available on “Preview” and “Run everywhere” refer to different deployment options. The “Preview” indicates a pre-release version for early access and testing, while “Run everywhere” implies a stable, production-ready version that can be deployed across various environments.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*hhQcmzn5RH-Oe4tJ.png" /></figure><p>Let us explore how the meta’s llama-3.1-nemotron-70b-instruct model looks like. Click on the Parameters tab to set the tokens, temperature, Frequency and Presence penalties.</p><p>Let’s explore the look and feel of Meta’s LLaMA-3.1-Nemotron-70B-Instruct model. To customize your experience, navigate to the <strong>Parameters</strong> tab to configure the settings Tokens, Temperature, Frequency Penalties, Presence Penalties</p><p>Simply paste your prompt into the Textbox and hit the arrow button. See the response for “How is the weather in New York city” prompt below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*iwmls-QwPdUBlCzv.png" /></figure><p><strong>Pro Tip:</strong> There is an Deploy tab for each of the models which allows to deploy the models in container using the option Linux with Docker which of course means you need the immense computing power but it can be also run on hosted endpoint and that means you are leaving the free territory as need to pay for running these models on hosted endpoints.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*2j3E1ts2VBnL0xWD.png" /></figure><p><strong>That’s It!</strong> You’re now all set to dive into the world of LLM Models, free of charge. Happy exploring.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dfe5f74435a4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Enable Smart Card authentication using YubiKey 5Ci security key on macOS.]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-link"><a href="https://learningisfun.medium.com/enable-smart-card-authentication-using-yubikey-5ci-security-key-on-macos-c524fe0b0726?source=rss-8e485a9bf87a------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://learningisfun.medium.com/enable-smart-card-authentication-using-yubikey-5ci-security-key-on-macos-c524fe0b0726?source=rss-8e485a9bf87a------2</link>
            <guid isPermaLink="false">https://medium.com/p/c524fe0b0726</guid>
            <category><![CDATA[apple]]></category>
            <category><![CDATA[security]]></category>
            <category><![CDATA[smart-card-readers]]></category>
            <category><![CDATA[yubikey]]></category>
            <category><![CDATA[mac]]></category>
            <dc:creator><![CDATA[Prashant]]></dc:creator>
            <pubDate>Sun, 26 Feb 2023 14:03:56 GMT</pubDate>
            <atom:updated>2025-08-09T01:50:49.996Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[How to prepare for (ISC)² Cloud Security Professional (CCSP) certification?]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-link"><a href="https://learningisfun.medium.com/how-to-prepare-for-isc-%C2%B2-cloud-security-professional-ccsp-certification-38062fd482f5?source=rss-8e485a9bf87a------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://learningisfun.medium.com/how-to-prepare-for-isc-%C2%B2-cloud-security-professional-ccsp-certification-38062fd482f5?source=rss-8e485a9bf87a------2</link>
            <guid isPermaLink="false">https://medium.com/p/38062fd482f5</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[cloud-security]]></category>
            <category><![CDATA[isc2]]></category>
            <category><![CDATA[pcc]]></category>
            <category><![CDATA[infosec]]></category>
            <dc:creator><![CDATA[Prashant]]></dc:creator>
            <pubDate>Mon, 05 Jul 2021 22:54:13 GMT</pubDate>
            <atom:updated>2025-08-09T01:50:24.537Z</atom:updated>
        </item>
    </channel>
</rss>