R/predict.filter.R

In sideChannelAttack: Side Channel Attack

predict.filter.NULL <-
function (object, newdata, ...) 
{
    return(newdata)
}
predict.filter.PCA <-
function (object, newdata, ...) 
{
    res = predict(object$mod, newdata)[,1:object$nbreVarX]
    return(res)
}
predict.filter.RegressionTreeFilter <-
function (object, newdata, ...) 
{
    if(is.matrix(newdata))
        nbreDExempleMax=dim(newdata)[1]
    else
        nbreDExempleMax=1
    resFinal=c()
    for(nbreDExemple in 1:nbreDExempleMax)
    {
        Nmax = object$nbreVarX
        a = newdata
        resume = TRUE
        shortResume = TRUE
        D = list()
        L = list()
        t = c(1:length(a))
        D[[1]] = t[1]
        D[[2]] = t[length(t)]
        L[[1]] = struct(g = integer(), d = integer())
        L[[1]]$g = t[1]
        L[[1]]$d = t[length(t)]
        ac = rep(mean(a), (t[length(t)]))
        Np = 1
        while (Np != Nmax) {
            if (is.struct(L[[1]])) {
                max = (L[[1]]$d - L[[1]]$g) * var(a[c(L[[1]]$g:L[[1]]$d)])
                ti = L[[1]]$g
                tj = L[[1]]$d
                LIndex = 1
                if (length(L) != 1) {
                    for (i in 2:length(L)) {
                      if (is.struct(L[[i]])) {
                        tmp = (L[[i]]$d - L[[i]]$g) * var(a[c(L[[i]]$g:L[[i]]$d)])
                        if (tmp > max) {
                          max = tmp
                          ti = L[[i]]$g
                          tj = L[[i]]$d
                          LIndex = i
                        }
                      }
                    }
                }
                L[[LIndex]] = NA
            }
            max = 0
            maxE = -1
            vartitj = (tj - ti) * var(a[which(t == ti):which(t == 
                tj)])
            for (e in which(t == ti):which(t == tj)) {
                varIE = var(a[ti:t[e]])
                if (is.na(varIE)) {
                    varIE = 0
                }
                varEJ = var(a[t[e]:tj])
                if (is.na(varEJ)) {
                    varEJ = 0
                }
                tmp = vartitj - (t[e] - ti) * varIE - (tj - t[e]) * 
                    varEJ
                if (tmp > max) {
                    max = tmp
                    maxE = e
                }
            }
            e = maxE
            te = t[e]
            distIE = length(which(t == ti):which(t == te))
            distEJ = length(which(t == te):which(t == tj))
            ac[ti:te] = rep(mean(a[ti:te]), distIE)
           ac[te:tj] = rep(mean(a[te:tj]), distEJ)
            Np = Np + 1
            trouve = FALSE
            trouveI = -1
            for (i in 1:length(L)) {
                if (!is.struct(L[[i]]) && !trouve) {
                    trouve = TRUE
                    trouveI = i
                }
            }
            if (!trouve) {
                trouveI = length(L) + 1
            }
            L[[trouveI]] = struct(g = integer(), d = integer())
            L[[trouveI]]$g = ti
            L[[trouveI]]$d = te
            trouve = FALSE
            trouveI = -1
            for (i in 1:length(L)) {
                if (!is.struct(L[[i]]) && !trouve) {
                    trouve = TRUE
                    trouveI = i
                }
            }
            if (!trouve) {
                trouveI = length(L) + 1
            }
            L[[trouveI]] = struct(g = integer(), d = integer())
            L[[trouveI]]$g = te
            L[[trouveI]]$d = tj
            D[[length(D) + 1]] = te
        }
        if (!resume) {
            resFinal = rbind(ac)
        }
        else {
            if (shortResume) {
                resFinal = rbind(unique(ac))
            }
            else {
                resTmp = unique(ac)
                res = resTmp
                for (i in resTmp) {
                    res = c(res, length(which(ac == i)))
                }
                resFinal = rbind(res)
            }
        }
    }
    return(resFinal)
}
predict.filter.mRMR <-
function (object, newdata, ...) 
{
    return(newdata[, object$filter])
}
predict.filter.MAX <-
function (object, newdata, ...) 
{
    if(is.matrix(newdata))
        nbreDExempleMax=dim(newdata)[1]
    else
        nbreDExempleMax=1
    res=c()
    for(nbreDExemple in 1:nbreDExempleMax)
    {
        res = rbind(res,sort(newdata[nbreDExemple,])[(dim(newdata)[2]-object$nbreVarX+1):(dim(newdata)[2])])
    }
    return(res)
}