-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathServerNew.m
220 lines (216 loc) · 8.02 KB
/
ServerNew.m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
% create a new server, called by main.m
classdef ServerNew
properties % (Access = private) % private variables
grid
clG
end
properties
n % arnold algorithm：number of rounds
a % arnold algorithm：
b % arnold algorithm：
end
methods
% ceate server objects
function obj = ServerNew()
obj.n = 5;
obj.a = 5;
obj.b = 6;
end
% Reduce the dimensions of the client. Receive information from the client and
% perform feature decomposition on the server to select appropriate feature vectors
function users = DP_PCA(obj, users)
numU = size(users,1);
[count, sumCol] = users{1,1}.getInfo_DPPCA1();
for i = 2:numU
[aa, bb] = users{i,1}.getInfo_DPPCA1();
count = count + aa;
sumCol = sumCol + bb;
end
meanCol = sumCol./count;
% When obtaining the three matrices H, B, and F for each client, consider using
% an image encryption algorithm (i.e. encrypting the matrices as images)
[encH, encB, encF, h] = users{1,1}.getInfo_DPPCA2(meanCol);
H = decrypt_arnold(obj, encH, h, h);
B = decrypt_arnold(obj, encB, 1, h);
F = decrypt_arnold(obj, encF, h, 1);
for i = 2:numU
[encH, encB, encF, h] = users{i,1}.getInfo_DPPCA2(meanCol);
H = H + decrypt_arnold(obj, encH, h, h);
B = B + decrypt_arnold(obj, encB, 1, h);
F = F + decrypt_arnold(obj, encF, h, 1);
end
% 4:
numer = count*H - B'*B;
deno = sqrt(count*F - B.^2) * sqrt(count*F - B.^2);
R = numer./deno;
% 5:
[W, lambda] = eig(R);
ev = abs((diag(lambda))');
[ev, rhoev] = sort(ev, 'descend');
W = W(:, rhoev);
% 6:
nn = length(ev);
kk = zeros(1, nn);
for i = 2:nn-1
kk(i) = (ev(i)-ev(i-1))/(ev(i+1)-ev(i));
end
lin = find(kk == max(kk));
meanEV = mean(ev(lin:end));
stdEV = std(ev(lin:end));
if ev(1) > meanEV+stdEV && lin > 2
W = W(:,1:lin-1);
else
for cut = 1:nn
if sum(ev(1:cut))/sum(ev) > 0.99
break;
end
end
W = W(:,1:cut);
end
% When transmitting reduced dimensional feature vectors from the
% server to the client, consider using encryption
W = real(W); % avoid the complex numbers
[h, w] = size(W);
encW = obj.encrypt_arnold(W);
for i = 1:numU
users{i,1} = users{i,1}.DPPCA(encW, h, w);
end
end
% use Arnold algorithm，encryption or chaotic
function newX = encrypt_arnold(obj, oldX)
[h, w] = size(oldX);
if h>w
max = h;
B = ones(max, max-w)*125;
img = cat(2, oldX, B);
else
max = w;
B = ones(max-h, max)*125;
img = cat(1, oldX, B);
end
[h, w] = size(img);
N = h;
newX = zeros(max, max);
for i = 1:obj.n
for y = 1:h
for x = 1:w
xx = mod((x-1)+obj.b*(y-1),N)+1;
yy = mod(obj.a*(x-1)+(obj.a*obj.b+1)*(y-1),N)+1;
newX(yy, xx) = img(y,x);
end
end
img = newX;
end
end
% using Arnold algorithm，decryption or recovery
function newX = decrypt_arnold(obj, oldX, s_h, s_w)
[h, w] = size(oldX);
newX = oldX;
N = h;
for i = 1:obj.n
for y = 1:h
for x = 1:w
xx = mod((obj.a*obj.b+1)*(x-1)-obj.b*(y-1),N)+1;
yy = mod(-obj.a*(x-1)+(y-1),N)+1;
oldX(yy, xx) = newX(y, x);
end
end
newX = oldX;
end
newX = imcrop(oldX,[0, 0, s_w, s_h]);
end
% normalize data
function users = Normalization(~, users)
numU = size(users,1);
[maxV, minV] = users{1,1}.getMaxMin();
for i = 2:numU
[maxI, minI] = users{i,1}.getMaxMin();
maxV = max(maxV, maxI);
minV = min(minV, minI);
end
for i = 1:numU
users{i,1} = users{i,1}.Normal(minV, maxV);
end
end
% calculate the divided grid points and the number of points within the grid
function [obj, L] = Meshing(obj, users, L_theta, n, theta, NCLUST)
numU = size(users,1);
for L = 2:n
gridT = users{1,1}.Partition(L);
for i = 2:numU
gridU = users{i,1}.Partition(L);
gridT = [gridT; gridU];
end
[obj.grid, ~, count] = unique(gridT, 'rows');
numG = sum(bsxfun(@eq, count, unique(count)'));
if length(numG)/n >= L_theta
break;
end
end
obj.clG = obj.SDC_DPC(numG, theta, L_theta, NCLUST);
end
% using SDC-DPC to label the grid
function clG = SDC_DPC(obj, numG, theta, L_theta, NCLUST)
Grid = obj.grid;
clG = zeros(size(Grid,1),1)-1;
K = ceil(sqrt(length(numG)));
distM = pdist2(Grid, Grid);
[sortM, ~] = sort(distM, 2);
rhoG = exp(-1.*mean(sortM(:,2:K+1),2));
numG = numG./max(numG);
rhoG = rhoG./max(rhoG);
rhoG = (1-L_theta)*numG' + L_theta*rhoG;
[~, ordG] = sort(rhoG, 'descend');
nneigh = zeros(size(Grid,1), 1);
delta = zeros(size(Grid,1), 1);
maxG = max(max(distM));
for i = 2:size(Grid,1)
delta(ordG(i)) = maxG;
for j = 1:i-1
if distM(ordG(i), ordG(j)) < delta(ordG(i))
delta(ordG(i)) = distM(ordG(i), ordG(j));
nneigh(ordG(i)) = ordG(j);
end
end
end
delta(ordG(1)) = maxG;
R = rhoG.*delta;
[~, ordR] = sort(R, 'descend');
iclG = [];
for i = 1:size(Grid,1)
if clG(ordR(i)) > 0
continue;
end
if length(iclG) == NCLUST
break;
end
Ci = ordR(i);
iclG = [iclG Ci];
clG(ordR(i)) = length(iclG);
neigh = find(distM(ordR(i),:)<2);
while ~isempty(neigh)
next = neigh(1);
neigh(1) = [];
if clG(next) > 0
continue;
end
if rhoG(next) > theta*rhoG(Ci)
clG(next) = length(iclG);
item = find(distM(next,:)<2);
neigh = [neigh item];
% reduce judgment
neigh = unique(neigh);
neigh(clG(neigh)>0) = [];
else
clG(next) = length(iclG);
end
end
end
for i = 1:length(ordG)
if clG(ordG(i)) < 0
clG(ordG(i)) = clG(nneigh(ordG(i)));
end
end
end
end
end