BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20240310T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20241103T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20240405T200100Z UID:80B9A84F-F299-45A7-8049-B0370AB1EC73 DTSTART;TZID=America/New_York:20240402T180000 DTEND;TZID=America/New_York:20240402T200000 DESCRIPTION:Overview\n\nThe field of Artificial Intelligence is taking the industry (and the world) by storm. The launch of ChatGPT started a mass ad option phenomenon\, which is now followed by an avalanche of new products and solutions addressing every possible problem space. Malicious hackers h ave also put AI to good use\, but so did the White Hat hackers\, such as F rédéric and his team.\n\nThis talk will level-set the understanding abou t Large Language Models (LLM) and Generative AI for computer scientists an d software developers who might not be intimately familiar with the field. Frédéric will then take us further into some of the applications in cyb ersecurity.\n\nAbstract\n\nLarge Language Models and Generative AI have co mpletely reshaped the landscape of Artificial Intelligence the last two ye ars. Progresses have been made on architectures\, training methods\, and t he community has shared large datasets along with pre-trained models\, all owing for new usages at a relatively low cost. In the cybersecurity defend er's path\, new malicious tools drove efforts on innovative methods to imp rove their detection. We propose a modern and effective method of detectin g file maliciousness\, by using an LLM initially trained on computer code. The focus will be on the process we set up and the decision we made to so lve this problem\, with an emphasize of the generalization of our approach . We will explain how LLM can help solve a large panel of problems related to texts.\n\nIn addition\, we will explain the outcomes of this work : in particular\, we will explain how we were able to use the trained model an d ask it "where it was wrong"\, shedding lights on errors in the training datasets. We will explain where these errors came from\, and how we were a ble to improve the model iteratively by correcting them\, allowing more pe ople to reproduce our findings and fixing pitfalls coming with noisy datas ets. Working in a field where sometimes the line is thin between an "admin istrator" and a "threat actor"\, we will also open the discussion on how y ou can define maliciousness depending on your objectives.\n\nKey takeaways :\n-What is an LLM and what it is not?\n-How much does it cost to use one for our own purpose?\n-What are the key steps of a training process?\n-An intro to cybersecurity : what are we trying to detect?\n-How did we build our datasets?\n-The results\n-What can the data engineer and the cybersec urity analyst learn from these results?\n\nCo-sponsored by: CIS 2024 | Use this IEEE Promo Code and enjoy 30% off your registration: IEEE_CIS2024\n\ nSpeaker(s): Frédéric Grelot\, Marc Lijour\n\nAgenda: \n6 pm Welcome and introductions by Marc Lijour\n\n6:30 pm - 7:30 pm Talk: "Demystifying AI and Large Language Models through a concrete use case" by Frédéric Grelo t\nFollowed by Q&As\n\n7:30 pm Networking\n\n8 pm The end\n\nRoom: ENGLG11 \, Bldg: George Vari Engineering and Computing Centre (ENG)\, 245 Church Street\, lower level \, Toronto\, Ontario\, Canada\, Virtual: https://even ts.vtools.ieee.org/m/412758 LOCATION:Room: ENGLG11 \, Bldg: George Vari Engineering and Computing Centr e (ENG)\, 245 Church Street\, lower level \, Toronto\, Ontario\, Canada\, Virtual: https://events.vtools.ieee.org/m/412758 ORGANIZER:marclijour@ieee.org SEQUENCE:32 SUMMARY:Large Language Models (AI) applications to Cybersecurity URL;VALUE=URI:https://events.vtools.ieee.org/m/412758 X-ALT-DESC:Description: <br /><p><strong>Overview</strong></p>\n<p>The fiel d of Artificial Intelligence is taking the industry (and the world) by sto rm. The launch of ChatGPT started a mass adoption phenomenon\, which is no w followed by an avalanche of new products and solutions addressing every possible problem space. Malicious hackers have also put AI to good use\, b ut so did the <em>White Hat</em> hackers\, such as Fr&eacute\;d&eacute\;ri c and his team.&nbsp\;</p>\n<p>This talk will level-set the understanding about Large Language Models (LLM) and Generative AI for computer scientist s and software developers who might not be intimately familiar with the fi eld. Fr&eacute\;d&eacute\;ric will then take us further into some of the a pplications in cybersecurity.</p>\n<p><strong>Abstract</strong>&nbsp\;</p> \n<p>Large Language Models and Generative AI have completely reshaped the landscape of Artificial Intelligence the last two years. Progresses have b een made on architectures\, training methods\, and the community has share d large datasets along with pre-trained models\, allowing for new usages a t a relatively low cost. In the cybersecurity defender's path\, new malici ous tools drove efforts on innovative methods to improve their detection. We propose a modern and effective method of detecting file maliciousness\, by using an LLM initially trained on computer code. The focus will be on the process we set up and the decision we made to solve this problem\, wit h an emphasize of the generalization of our approach. We will explain how LLM can help solve a large panel of problems related to texts.</p>\n<p><br >In addition\, we will explain the outcomes of this work : in particular\, we will explain how we were able to use the trained model and ask it "whe re it was wrong"\, shedding lights on errors in the training datasets. We will explain where these errors came from\, and how we were able to improv e the model iteratively by correcting them\, allowing more people to repro duce our findings and fixing pitfalls coming with noisy datasets. Working in a field where sometimes the line is thin between an "administrator" and a "threat actor"\, we will also open the discussion on how you can define maliciousness depending on your objectives.<br><br><u>Key takeaways :</u> <br>-What is an LLM and what it is not?<br>-How much does it cost to use o ne for our own purpose?<br>-What are the key steps of a training process?< br>-An intro to cybersecurity : what are we trying to detect?<br>-How did we build our datasets?<br>-The results<br>-What can the data engineer and the cybersecurity analyst learn from these results?</p><br /><br />Agenda: <br /><p>6 pm&nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbs p\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\;Welcome and introductio ns by Marc Lijour</p>\n<p>6:30 pm - 7:30 pm&nbsp\; &nbsp\; Talk: "Demystif ying AI and Large Language Models through a concrete use case" by Fr&eacut e\;d&eacute\;ric Grelot&nbsp\;<br>&nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &n bsp\; &nbsp\; &nbsp\; &nbsp\; Followed by Q&amp\;As</p>\n<p>7:30 pm&nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; & nbsp\;Networking</p>\n<p>8 pm&nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp \; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\; &nbsp\;The end< /p> END:VEVENT END:VCALENDAR