Recently, the first “Huawei Cup” Postgraduate Network Security Innovation Competition in Chinawas held.Wuhan University (WHU)won five first prizes, ranking first among universities in China. The team called LuojiaHill’sAIMM fromtheSchool ofComputerScience and Engineering tookthenational first prizein thecreative workscompetitionfor their outstanding project“InvisDefense invisibility cloak, a wearable and all-weather equipment defending malicious detection of intelligent cameras”.The teamwas supervisedby Professor Wang Zheng fromtheSchool of Computer Scienceandconsistedof doctoral student Wei Hui fromtheSchool of Computer Science, postgraduatesLi Zhuboand Dai ShuyvfromtheSchool of Cyber Science and Engineering, and postgraduate Jian Zehua fromtheEconomics and Management School.
Team members of LuojiaHill’sAIMM
Unlikestealth fighterjetsthat are just invisible to radar,theInvisDefense invisibility cloak makes it invisible tomachine recognition,not the naked eye.“Visual neural network refers to machine vision, not visual recognition of humans.It is widely used invideo surveillance equipment withthepedestrian detection functionandsmart cars,” explainedProf.Wang.
TheInvisDefense invisibility cloakinvolvesa kind of camouflage patterndesignedby a new algorithm, invalidating the commonlyused method of pedestrian detection. “In layman’s terms, it means camerascan detect you but cannotdetermine thatyouarehuman.”
Principle analysis and effect displays
TheInvisDefense invisibility cloakcanprovideall-weatherstealthdefense. Inthedaytime, particular camouflage patterns disturbthevisual neural network effectivelytomakeit blind. Inthenighttime, however, InvisDefense embeds temperature control materials, thus realizing stealth under thermal infrared imagingthroughspecial color blocks. As a brand-new defense product,theInvisDefense’sinvisibility cloak can be invisible under machine visionandthereforebeused to protectpersonalprivacy andsecurity.
Different models in daytime and nighttime
Wei Huiwas in charge ofthecore algorithm and designin the team. Since 2021, he has been doingresearchontrusted AI. In August, 2022,whenthethemesof thecompetitionwereannounced,hefound thatthecybersecuritytopicwas in accord with his research interests. Therefore, he teamedupwith other members and began to prepare for thecompetition.After the team was formed, theyimmediately began topreparefor the competition.In the beginning, how to transformtheir theoreticalachievements intoproductsprovedquitedifficult,so a corethemeneededtobe determined. Afterdiscussionsand compromiseswithinthe team, the product—invisibility cloak-was jointlyagreed.
The slides forthecompetition presentation
Turningtheory into realityis not easy. “The first problem is toachievea balance. For example, youmay fool thecameras if youhideyourself with color blocks, butit’s hardto escape from human eyes,forit’ll make you conspicuous. As a result, wehadtodesigntheleast attractive patternsthatare able tonullifythe effect of computer vision bythealgorithm.” To realize the balance, memberskept going throughcomputersimulations,experiencingabout 700 failures. “Sometimes, we printedapatterntotestitin a realisticenvironmentandagain found that it didn’t work, so we had to continue to improve it.”Wei Huisaidthat by cooperating and encouragingeach other, members overcameconsiderablepressure and, in the end,tested the ideal results.
Algorithm framework of InvisDefense
“We trieditout on campus,andfound thatInvisDefense is effectiveboth inthedaytime andatnighttime.At present,we can achieve a 57 percentreductionpedestrian detection accuracy,which can be higherin the future.” When peoplewearinganInvisDefenseinvisibility cloakand thosewearingordinary clothesare movinginthesame scene, monitoring devices will only captureand framepeople who are not wearingtheinvisibilitycloak as“person”.
Test effect oftheinvisibility cloak
Another challengein this competitionwas to make InvisDefenseoperateforas long as possible. The team aimedto find out the most economical plan. They first testedthewarmer pad and cooling sheet. After changing partial temperature, they neededto figure out an arrangement tosuccessfullydisturb computer visual recognition. Through their unremitting efforts, the best plan came into beingbased on the idea thatcomputers’ infrared imaging recognition systems will go blind iffour pieces of irregular-shaped temperature control materialsare embedded.
After winning the prize, members wereeagerto express their sincere gratitude to their tutor,Prof.Wang.Whethertheprogrammetabottleneckortheyhad todiscusshow topresentthe best results,Prof.Wangalways stoodby their side,givingthe teamscientificguidance, andsupportingthem.
Mutual encouragementbetween team members wasalso powerful. Wei Hui was selected to dothepresentationin the final competition.The time for preparation was quitelimited for it onlylastedhalf a monthbeforethefinal stage. WeiHui frankly admittedtohispoorexpressiveabilityand said that itwashis teammates’ positive encouragement andsuggestionsthatgave himtheconfidenceto defend.
Members prepareunder the guidance ofProf.Wang
Fromthepreliminary tothefinal competition,thefour membersfought together for their common purpose. Every day and nightthatthey wrotereportsand tested resultstogetherwill definitely becomeone of theirunforgettable memories.
At present,theInvisDefense invisibility cloakhasreceived two patents, andtherelatedresultshavealsobeenaccepted byAAAI2023,atopacademicconferenceonartificialintelligencein category A of CCF (Chinese Computer Society). All the reviewersspokehighlyoftheirprojectasbeinginteresting, cost-effective,havinghighoperability and excellent results.
Materials and results
Talking about thepractical applicationprospectsof the project,Wang Zheng believes thattheinvisibility cloak will likely contributeto the national defense of China. For the civilian side, the value of the project is to reverse the promotion ofthe security ofartificial intelligence(AI)models. “Researching Internet security isabout discoveringflaws and thenpluggingloopholes through constant confrontationbetweenattacks and defense. Our results prove that both AI and computer recognition technologies have room for further development,”saidProf.Wang. “The research for theprogramis connected with many principles inAIfields. For example, by doing analogous tests, intelligent driving will identify road conditions moreaccurately,” Prof.Wang added.
Further research and applications oftheInvisDefenseprogram
When it comes tothe feelings oftheparticipantsinthiscompetition,the teammembersthoughtthat this opportunityhaseffectivelycultivatedtheirscientific innovation and engineeringpracticeability.Italsoencouragedmore students toactivelytrytheirfavoritedirectionandparticipatein relatedcompetitionsduring theircollegeyears. AsProf.Wangsaid, “Doing research isfull ofpleasure.What matters most is that students havethechance to choose theirfavoritedirectionand serve all walks of life.”
Final of“Huawei Cup” Postgraduate Network Security Innovation Competition
At present, the teamcontinuesto research stealth defensesystemsaiming at machine vision, with visible light, thermal infrared, remote sensing and other imaging, including stealth ofpersons, cars, and other things. In the future,the teamwillcontinueto improvetheirprogramInvisDefense invisibility cloak andconductfurthervaluable research inthefield ofAIsecurity.
Just like Wei Hui and his teammates, moreandmore WHU studentsfollow a similar route. Byofferingand developingtheir expertise,they create, innovate, andcontinuouslypracticetofurther improvethe development of various fields.
Rewritten by Gao Fei
Edited byLi Jing,Su xinyue, Sylvia, Xi Bingqing