forced pdf json strings to be utf-8 or base64 encoded

libclamav/pdf.c

@@ -3511,22 +3511,86 @@ static void pdf_export_json(struct pdf_struct *pdf)
         goto cleanup;
     }
 
-    if (pdf->stats.author)
+    if (pdf->stats.author) {
+        if (cli_isutf8(pdf->stats.author, strlen(pdf->stats.author)))
             cli_jsonstr(pdfobj, "Author", pdf->stats.author);
-    if (pdf->stats.creator)
+        else {
+            char *b64 = (char *)cl_base64_encode(pdf->stats.author, strlen(pdf->stats.author));
+            cli_jsonstr(pdfobj, "Author", b64);
+            cli_jsonbool(pdfobj, "Author_base64", 1);
+            free(b64);
+        }
+    }
+    if (pdf->stats.creator) {
+        if (cli_isutf8(pdf->stats.creator, strlen(pdf->stats.creator)))
             cli_jsonstr(pdfobj, "Creator", pdf->stats.creator);
-    if (pdf->stats.producer)
+        else {
+            char *b64 = (char *)cl_base64_encode(pdf->stats.creator, strlen(pdf->stats.creator));
+            cli_jsonstr(pdfobj, "Creator", b64);
+            cli_jsonbool(pdfobj, "Creator_base64", 1);
+            free(b64);
+        }
+    }
+    if (pdf->stats.producer) {
+        if (cli_isutf8(pdf->stats.producer, strlen(pdf->stats.producer)))
             cli_jsonstr(pdfobj, "Producer", pdf->stats.producer);
-    if (pdf->stats.modificationdate)
+        else {
+            char *b64 = (char *)cl_base64_encode(pdf->stats.producer, strlen(pdf->stats.producer));
+            cli_jsonstr(pdfobj, "Producer", b64);
+            cli_jsonbool(pdfobj, "Producer_base64", 1);
+            free(b64);
+        }
+    }
+    if (pdf->stats.modificationdate) {
+        if (cli_isutf8(pdf->stats.modificationdate, strlen(pdf->stats.modificationdate)))
             cli_jsonstr(pdfobj, "ModificationDate", pdf->stats.modificationdate);
-    if (pdf->stats.creationdate)
+        else {
+            char *b64 = (char *)cl_base64_encode(pdf->stats.modificationdate, strlen(pdf->stats.modificationdate));
+            cli_jsonstr(pdfobj, "ModificationDate", b64);
+            cli_jsonbool(pdfobj, "ModificationDate_base64", 1);
+            free(b64);
+        }
+    }
+    if (pdf->stats.creationdate) {
+        if (cli_isutf8(pdf->stats.creationdate, strlen(pdf->stats.creationdate)))
             cli_jsonstr(pdfobj, "CreationDate", pdf->stats.creationdate);
-    if (pdf->stats.title)
+        else {
+            char *b64 = (char *)cl_base64_encode(pdf->stats.creationdate, strlen(pdf->stats.creationdate));
+            cli_jsonstr(pdfobj, "CreationDate", b64);
+            cli_jsonbool(pdfobj, "CreationDate_base64", 1);
+            free(b64);
+        }
+    }
+    if (pdf->stats.title) {
+        if (cli_isutf8(pdf->stats.title, strlen(pdf->stats.title)))
             cli_jsonstr(pdfobj, "Title", pdf->stats.title);
-    if (pdf->stats.subject)
+        else {
+            char *b64 = (char *)cl_base64_encode(pdf->stats.title, strlen(pdf->stats.title));
+            cli_jsonstr(pdfobj, "Title", b64);
+            cli_jsonbool(pdfobj, "Title_base64", 1);
+            free(b64);
+        }
+    }
+    if (pdf->stats.subject) {
+        if (cli_isutf8(pdf->stats.subject, strlen(pdf->stats.subject)))
             cli_jsonstr(pdfobj, "Subject", pdf->stats.subject);
-    if (pdf->stats.keywords)
+        else {
+            char *b64 = (char *)cl_base64_encode(pdf->stats.subject, strlen(pdf->stats.subject));
+            cli_jsonstr(pdfobj, "Subject", b64);
+            cli_jsonbool(pdfobj, "Subject_base64", 1);
+            free(b64);
+        }
+    }
+    if (pdf->stats.keywords) {
+        if (cli_isutf8(pdf->stats.keywords, strlen(pdf->stats.keywords)))
             cli_jsonstr(pdfobj, "Keywords", pdf->stats.keywords);
+        else {
+            char *b64 = (char *)cl_base64_encode(pdf->stats.keywords, strlen(pdf->stats.keywords));
+            cli_jsonstr(pdfobj, "Keywords", b64);
+            cli_jsonbool(pdfobj, "Keywords_base64", 1);
+            free(b64);
+        }
+    }
     if (pdf->stats.ninvalidobjs)
         cli_jsonint(pdfobj, "InvalidObjectCount", pdf->stats.ninvalidobjs);
     if (pdf->stats.njs)

libclamav/str.c

@@ -666,3 +666,49 @@ char *cli_utf16_to_utf8(const char *utf16, size_t length, utf16_type type)
     s2[j] = '\0';
     return s2;
 }
+
+int cli_isutf8(const unsigned char *buf, unsigned int len)
+{
+	unsigned int i, j;
+
+    for(i = 0; i < len; i++) {
+        if((buf[i] & 0x80) == 0) {  /* 0xxxxxxx is plain ASCII */
+            continue;
+        } else if((buf[i] & 0x40) == 0) { /* 10xxxxxx never 1st byte */
+            return 0;
+        } else {
+            unsigned int following;
+
+            if((buf[i] & 0x20) == 0) {		/* 110xxxxx */
+                /* c = buf[i] & 0x1f; */
+                following = 1;
+            } else if((buf[i] & 0x10) == 0) {	/* 1110xxxx */
+                /* c = buf[i] & 0x0f; */
+                following = 2;
+            } else if((buf[i] & 0x08) == 0) {	/* 11110xxx */
+                /* c = buf[i] & 0x07; */
+                following = 3;
+            } else if((buf[i] & 0x04) == 0) {	/* 111110xx */
+                /* c = buf[i] & 0x03; */
+                following = 4;
+            } else if((buf[i] & 0x02) == 0) {	/* 1111110x */
+                /* c = buf[i] & 0x01; */
+                following = 5;
+            } else {
+                return 0;
+            }
+
+            for(j = 0; j < following; j++) {
+                if(++i >= len)
+                    return 0;
+
+                if((buf[i] & 0x80) == 0 || (buf[i] & 0x40))
+                    return 0;
+
+                /* c = (c << 6) + (buf[i] & 0x3f); */
+            }
+        }
+    }
+
+    return 1;
+}

libclamav/str.h

@@ -58,5 +58,7 @@ typedef enum {
 } utf16_type;
 char *cli_utf16_to_utf8(const char *utf16, size_t length, utf16_type type);
 
+int cli_isutf8(const unsigned char *buf, unsigned int len);
+
 size_t cli_strlcat(char *dst, const char *src, size_t sz); /* libclamav/strlcat.c */
 #endif