Cibiyar Sadarwar Neuronal na Graph Wanda Ya Inganta Gano Rashin Tsaro a cikin Kwangilar Smart na Blockchain na Ilimi

Teburin Abubuwan Ciki

1. Gabatarwa

Blockchain na Ilimi yana wakiltar aikace-aikacen fasahar blockchain don canza tsarin ilimi na al'ada. Bayyananniyar bayyanawa da halayen rashin canzawa na blockchain sun sa ya dace musamman don sarrafa ƙimar ɗalibi, takaddun shaida na ilimi, da haɗin gwiwar masana'antu da ilimi. Tare da ci gaban fasahar kwangilar Smart na Ethereum, cibiyoyin ilimi na iya gina tsarin ciniki mai hankali da dandamali na koyo waɗanda ke aiwatar da kai tsaye lokacin da an cika sharuɗɗan da aka ƙayyade.

Duk da haka, rashin canzawar blockchain yana gabatar da manyan ƙalubalen tsaro. Da zarar an tura su, kwangilolin Smart ba za a iya gyara su ba, wanda haka ya sa gano raunin ya zama mahimmanci kafin a tura su. Wannan binciken yana magana da buƙatar mahimmanci don ingantaccen gano raunin a cikin kwangilolin Smart na blockchain na ilimi ta amfani da Cibiyar Sadarwar Neuronal na Graph (GNNs).

2. Hanyar Aiki

2.1 Rarraba Bytecode

Hanyar da aka tsara ta fara ne da rarraba bytecode ɗin kwangilar Smart na Ethereum don samun lambobin aiki (opcodes). Wannan tsari ya haɗa da canza ƙananan bytecode ɗin zuwa jerin opcode masu iya karantawa da mutum waɗanda ke adana ainihin dabaru na kwangila yayin da suke ba da damar binciken tsari.

2.2 Gina Taswirar Kula da Kwarara

Ana gano tubalan asali daga jerin opcode, kuma ana ƙara gefuna tsakanin tubalan bisa ga dabaru na aiwatarwa. Taswirar Kula da Kwarara (CFG) da aka samu tana ɗaukar hanyoyin aiwatarwa na shirin da abubuwan dogaro na sarrafawa, tana ba da wakilcin tsarin da ya dace don bincike na tushen zane.

2.3 Tsarin Ginin Model na GNN

Model ɗin GNN yana sarrafa CFG don gano raunin. Tsarin ginin yana amfani da yadudduka na haɗin gwiwar graph waɗanda ke tattara bayanai daga nodes maƙwabta, wanda ke ba da damar model ɗin koyon alamu masu nuna raunin tsaro a faɗin tsarin kula da kwarara na kwangila.

3. Aiwarta na Fasaha

3.1 Tsarin Lissafi

Ana iya wakiltar aikin GNN ta hanyar lissafi ta amfani da dabarar jujjuyawar graph:

$H^{(l+1)} = \sigma(\tilde{D}^{-\frac{1}{2}}\tilde{A}\tilde{D}^{-\frac{1}{2}}H^{(l)}W^{(l)})$

inda $\tilde{A} = A + I$ shine matrix na kusanci tare da haɗin kai, $\tilde{D}$ shine matrix na digiri, $H^{(l)}$ yana wakiltar siffofi na node a Layer $l$, $W^{(l)}$ ma'auni ne masu koyarwa, kuma $\sigma$ shine aikin kunnawa.

3.2 Aiwartar Code

class SmartContractGNN(nn.Module):
    def __init__(self, input_dim, hidden_dim, output_dim):
        super(SmartContractGNN, self).__init__()
        self.conv1 = GCNConv(input_dim, hidden_dim)
        self.conv2 = GCNConv(hidden_dim, hidden_dim)
        self.classifier = nn.Linear(hidden_dim, output_dim)
        
    def forward(self, x, edge_index):
        # Graph convolution layers
        x = F.relu(self.conv1(x, edge_index))
        x = F.dropout(x, training=self.training)
        x = self.conv2(x, edge_index)
        
        # Global mean pooling
        x = global_mean_pool(x, batch=None)
        
        # Classification
        return self.classifier(x)

4. Sakamakon Gwaji

Ƙimar gwajin ta nuna cewa hanyar da aka tsara na tushen GNN ta cimma ingantaccen gano raunin tare da ƙarancin yadudduka na haɗin gwiwar graph idan aka kwatanta da hanyoyin al'ada. Model ɗin ya nuna ƙarfi na musamman wajen gano raunin dogaro da alamar lokaci, waɗanda ke da mahimmanci a cikin aikace-aikacen blockchain na ilimi inda ayyuka masu mahimmanci na lokaci ke sarrafa samun damar albarkatun ilimi da takaddun shaida.

Sakamakon ya nuna cewa haɗin binciken bytecode na kwangila da samfuran GCN yana ba da ingantaccen gano raunin, tare da samfurin da ya cimma ingantaccen daidaito yayin kiyaye ingantaccen lissafi. Hanyar ta yi nasarar gano alamu masu rauni a cikin taswirar kula da kwarara waɗanda kayan aikin bincike na tsaye na al'ada za su iya rasa.

5. Bincike da Tattaunawa

Wannan binciken ya gabatar da wani gagarumin ci gaba a cikin tsaron kwangilar Smart don aikace-aikacen blockchain na ilimi. Haɗa Cibiyar Sadarwar Neuronal na Graph tare da binciken bytecode na al'ada yana wakiltar wata sabuwar hanya wacce ke magance ƙalubalen musamman da rashin canzawar blockchain ya haifar. Ba kamar hanyoyin al'ada waɗanda suka dogara da daidaita tsari ko aiwatar da alama ba, hanyar tushen GNN tana koyon tsarin tsarin raunin kai tsaye daga taswirar kula da kwarara.

Gudunmawar fasaha ta ta'allaka ne a cikin nuna cewa tsarin ginin GNN mara zurfi zai iya ɗaukar rikitattun alaƙa a cikin code ɗin kwangilar Smart yadda ya kamata, yana ƙalubalantar sanin al'ada cewa cibiyoyin sadarwa masu zurfi sun zama dole don rikitarwar gane tsari. Wannan binciken ya yi daidai da bincike na kwanan nan a cikin koyon wakilcin graph, kamar aikin Kipf da Welling (2017) akan rarrabuwa na rabin-supervision tare da cibiyoyin sadarwar graph, wanda ya nuna cewa sauƙaƙan tsarin ginin na'ura mai kwakwalwa na iya cimma sakamako na zamani akan bayanan da suka tsara zane.

Idan aka kwatanta da kayan aikin bincike na kwangilar Smart na al'ada kamar Oyente ko Mythril, waɗanda da farko ke amfani da aiwatar da alama da binciken gurɓataccen abu, hanyar GNN tana ba da fa'idodi da yawa. Tana iya koyo daga gaba ɗaya tsarin kula da kwarara maimakon dogaro da ƙayyadaddun tsarin rauni, yana mai da shi mafi dacewa da sabbin nau'ikan rauni. Wannan iyawa yana da matukar muhimmanci a cikin saurin canzawar fagen barazanar tsaro na blockchain.

Mayar da hankali kan aikace-aikacen blockchain na ilimi yana da dacewa, ganin ƙaruwar amfani da fasahar blockchain a cikin takaddun shaida na ilimi da tsarin sarrafa koyo. Kamar yadda aka lura a cikin ma'auni na IEEE Blockchain a cikin Ilimi, raunin tsaro a cikin waɗannan tsare-tsare na iya haifar da sakamako mai nisa, suna lalata amincin bayanan ilimi da takaddun shaida. Hanyar da aka bayyana a cikin wannan takarda tana magance waɗannan damuwa ta hanyar samar da ingantaccen hanya don gano raunin kafin turawa.

Duk da haka, binciken kuma ya nuna buƙatar manyan, tarin bayanai daban-daban na kwangilolin Smart masu rauni don horarwa. Aikin gaba zai iya amfana daga haɗin gwiwa tare da ƙungiyoyi kamar Cibiyar Ƙididdiga da Fasaha ta Ƙasa (NIST) don haɓaka daidaitattun bayanan rauni don binciken tsaro na blockchain.

6. Aikace-aikace na Gaba

Hanyar da aka tsara tana da babban yuwuwar fa'ida mai faɗi fiye da blockchain na ilimi. Hanyoyin gaba sun haɗa da:

• Gano Rashin Tsaro na Ketare Dandamali: Miƙa hanyar zuwa wasu dandamali na blockchain kamar Hyperledger da Corda
• Sa ido na Ainihi: Haɓaka tsare-tsare don ci gaba da tantance raunin kwangilolin da aka tura
• Ƙirƙirar Faci ta kai tsaye: Haɗawa da tsarin AI don ba da shawarar gyaran rauni
• Haɗa Kayan Aikin Ilimi: Haɗa tsarin gano cikin manhajojin haɓaka blockchain

7. Bayanan Kafa

1. Z. Wang et al., "Cibiyar Sadarwar Neuronal na Graph don Gano Rashin Tsaro na Kwangilar Smart," Jaridar Binciken Blockchain, 2023.
2. T. N. Kipf da M. Welling, "Rarrabuwa na Rabin-Supervision tare da Cibiyar Sadarwar Graph," ICLR, 2017.
3. L. Luu et al., "Sanya Kwangilolin Smart Su Zama Masu Hankali," CCS 2016.
4. Ma'auni na IEEE don Blockchain a cikin Ilimi, IEEE Std 2418.1-2020.
5. A. M. Antonopoulos da G. Wood, "Ƙwarewa akan Ethereum: Gina Kwangilolin Smart da DApps," O'Reilly Media, 2018.
6. Cibiyar Ƙididdiga da Fasaha ta Ƙasa, "Bayanin Fasahar Blockchain," NISTIR 8202, 2018.