{"677959":{"#nid":"677959","#data":{"type":"news","title":"The Sherlock Holmes of AI ","body":[{"value":"\u003Cp\u003EImagine being a passenger in a self-driving car as the vehicle starts veering off the road. It\u2019s not a faulty sensor causing the dangerous situation\u0026nbsp;\u2014 it\u2019s a cyberattack. Hackers can access the deep learning (DL) neural networks at the heart of the vehicle\u2019s computer system, compromising the safety of its passengers, as well as other drivers and pedestrians.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EStopping such cyberattacks requires understanding them first, but this can be challenging. Finding a computing system\u2019s exact deep neural network has many roadblocks. They are often proprietary and, therefore, inaccessible to investigators without considerable legal intervention. Another common problem is that they are updated frequently, making it difficult for investigating researchers to access the most current network iteration. But a new tool from Georgia Tech could unlock the mysterious malware on myriad neural networks in everything from self-driving cars to the IMDB entertainment database. AI Psychiatry (AiP) is a postmortem cybersecurity forensic tool that uses artificial intelligence to recover the exact models a compromised machine runs on and discover where the fatal error occurred.\u003C\/p\u003E\u003Cp\u003E\u201cWe trust self-driving cars with our lives and ChatGPT with our careers, but when those systems fail, how are we going to investigate them?\u201d said\u0026nbsp;\u003Ca href=\u0022https:\/\/saltaformaggio.ece.gatech.edu\/\u0022\u003EBrendan Saltaformaggio\u003C\/a\u003E, an associate professor with joint appointments in the School of\u0026nbsp;\u003Ca href=\u0022https:\/\/scp.cc.gatech.edu\/\u0022\u003E\u0026nbsp;Cybersecurity and Privacy\u003C\/a\u003E\u0026nbsp;and the School of\u0026nbsp;\u003Ca href=\u0022https:\/\/www.ece.gatech.edu\/\u0022\u003EElectrical and Computer Engineering\u003C\/a\u003E (ECE).\u003C\/p\u003E\u003Cp\u003EAiP can recover the original DL model on both the local network\u2019s memory and the graphics processing unit that trains the network. It can accomplish this without any specific knowledge of the model\u2019s framework, platform, or version. Instead, it recreates the model using what \u003Ca\u003ESaltaformaggio\u003C\/a\u003E refers to as \u201cclues,\u201d or common components in all neural networks. These include weights, biases, shapes, and layers from the model\u2019s memory image \u2014 a frozen set of the bits and bytes operating when the model is running normally. The memory image is crucial because it enables AiP to compare it with the model post-attack.\u003C\/p\u003E\u003Cp\u003E\u201cThese models often refine their information as they go, based on their current environment, so an attack might happen as a result of an attacker poisoning the information a particular model is learning,\u201d said David Oygenblik, an ECE Ph.D. student. \u201cWe determined that a memory image would capture all those changes that occur during a runtime.\u201d\\\u003C\/p\u003E\u003Cp\u003EOnce the model is recovered, AiP can run it on another device, letting investigators test it thoroughly to determine where the flaws lie. AiP has been tested with different versions of both popular machine learning frameworks (TensorFlow and PyTorch) and datasets (CIFAR-10, LISA, and IMDB). It successfully recovered and rehosted 30 models with 100% accuracy.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u201cBefore our research, you couldn\u0027t go to the cyber \u2018crime scene\u2019 and find clues because there was no technique available to do that,\u201d Saltaformaggio said. \u201cThat\u0027s what we are pioneering in the cyber forensics lab right now \u2014 techniques to get that evidence out of a crime scene.\u201d\u003C\/p\u003E\u003Cp\u003ETools like AiP will allow cyber investigators to see the whole picture immediately. Solving cybercrimes can help prevent future ones, from safeguarding a user\u2019s data to keeping a car on the road.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u003Ca href=\u0022https:\/\/ece.gatech.edu\/news\/2023\/12\/gtri-georgia-tech-develop-ai-psychiatry-advance-national-security\u0022\u003E\u003Cem\u003EAiP\u003C\/em\u003E\u003C\/a\u003E\u003Cem\u003E is the inaugural winner\u0026nbsp;\u003C\/em\u003E\u003Ca href=\u0022https:\/\/www.gtri.gatech.edu\/newsroom\/gtri-graduate-student-research-fellowship-program-continues-expand-third-year\u0022\u003E\u003Cem\u003Eof GTRI\u0027s Graduate Student Fellowship Program\u003C\/em\u003E\u003C\/a\u003E\u003Cem\u003E.\u0026nbsp;\u003C\/em\u003E\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003ENew tool AI Psychiatry recovers compromised deep-learning models so researchers can understand what went wrong.\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"New tool AI Psychiatry recovers compromised deep-learning models so researchers can understand what went wrong."}],"uid":"34541","created_gmt":"2024-10-29 14:21:13","changed_gmt":"2024-11-04 13:47:37","author":"Tess Malone","boilerplate_text":"","field_publication":"","field_article_url":"","dateline":{"date":"2024-10-29T00:00:00-04:00","iso_date":"2024-10-29T00:00:00-04:00","tz":"America\/New_York"},"extras":[],"hg_media":{"675455":{"id":"675455","type":"image","title":"GettyImages-1902996278.jpg","body":"\u003Cp\u003ECourtesy of Getty Images\u003C\/p\u003E","created":"1730211995","gmt_created":"2024-10-29 14:26:35","changed":"1730211995","gmt_changed":"2024-10-29 14:26:35","alt":"Self-driving car","file":{"fid":"259082","name":"GettyImages-1902996278.jpg","image_path":"\/sites\/default\/files\/2024\/10\/29\/GettyImages-1902996278.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2024\/10\/29\/GettyImages-1902996278.jpg","mime":"image\/jpeg","size":8809887,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2024\/10\/29\/GettyImages-1902996278.jpg?itok=iJyNJ0xC"}}},"media_ids":["675455"],"groups":[{"id":"1214","name":"News Room"},{"id":"1188","name":"Research Horizons"}],"categories":[],"keywords":[{"id":"187915","name":"go-researchnews"}],"core_research_areas":[],"news_room_topics":[{"id":"71881","name":"Science and Technology"}],"event_categories":[],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp\u003ETess Malone, Senior Research Writer\/Editor\u0026nbsp;\u003C\/p\u003E\u003Cp\u003Etess.malone@gatech.edu\u003C\/p\u003E","format":"limited_html"}],"email":[],"slides":[],"orientation":[],"userdata":""}}}